A message selection system for network environments.
The message selection system addresses message fatigue by using confidence values and cooldown factors to deliver personalized messages at optimal times, enhancing user engagement and resource efficiency.
Patent Information
- Application Number
- JP2021562078
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-04-17
- Filing Date
- 2020-04-16
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2040-04-16
AI Technical Summary
Users of customized messaging systems experience message fatigue, leading to reduced effectiveness and wastage of computational resources due to messages not being acted upon at the client device, which is a challenge for applications requiring user interaction.
A message selection system that uses confidence values and cooldown factors to selectively send messages based on user preferences, application engagement, and response history, adjusting message delivery timing to reduce fatigue and enhance user interaction.
The system increases user engagement and application effectiveness by delivering personalized messages at optimal times, reducing message fatigue and optimizing resource usage.
Smart Images

Figure 0007581237000003 
Figure 0007581237000004 
Figure 0007581237000005
Abstract
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application claims priority to U.S. Provisional Patent Application No. 62 / 835,267, filed April 17, 2019, entitled “MESSAGE SELECTION AND TRANSMISSION BASED ON CONFIDENCE VALUES AND COOL DOWN FACTORS IN A NETWORKED ENVIRONMENT,” and U.S. Provisional Patent Application No. 62 / 835,271, filed April 17, 2019, entitled “INDIRECT UPDATES TO MESSAGING SYSTEMS IN A NETWORKED ENVIRONMENT,” both of which are incorporated by reference in their entireties herein. [Background technology]
[0002] Users of customized messaging systems can suffer from message fatigue, which reduces the effectiveness of messages to their intended recipients. Message fatigue can lead to wasted computational resources and bandwidth because messages sent over the network to a user's client device are not acted upon at the client device. For applications with desired user interaction and response, customized messaging can be a tool to achieve user engagement goals. Summary of the Invention
[0003] This disclosure describes systems and methods for selecting and sending messages based on a trust value and a cooldown factor. The messaging system can send relevant messages to users at relevant times. The personalized messaging system can increase user engagement and application effectiveness. Appropriate content delivered at the right time for each user can positively impact a user's engagement with the application. Personalized and relevant messages delivered to each user in a timely manner via the application can lead to real actions taken by the user to improve their health, happiness, fitness, and overall quality of life.
[0004] To address some of the challenges associated with message fatigue and other related challenges, the present disclosure relates to a message selection system that can invoke multiple message objects that can generate multiple messages when invoked at any time. A confidence value can be generated for each of the messages, and the confidence value associated with each message can be used to select one or more messages to send to an application on a client device. The message selection system can adjust the confidence value associated with each selected message based on a cool-down factor. The message selection system can use the cool-down factor to adjust the confidence value of each of the selected messages based on various factors, including, but not limited to, the amount of time since the last message was received by the client device, the total number of messages that a user of the client device has selected to receive within a given time period, and the amount of time that a user of the client device has been awake or active. The message selection system can then determine to send one or more messages selected based on the adjusted confidence value based on the adjusted confidence value of each selected message.
[0005] The message selection system may select messages based on the likelihood that the messages will achieve a desired endpoint or goal, while adjusting the likelihood based on the time since the last message was delivered to the client device, to reduce the likelihood that the user will experience message fatigue. The message selection system may reduce the likelihood that the user will experience message fatigue by selectively sending or dispatching messages to the user's client device based on the likelihood that the messages will have the intended effect. Because each user experiences message fatigue with a different number and frequency of messages, a framework or model may be constructed to deliver personalized messages at the appropriate time, incorporating information about each user's message preferences, application engagement, and responses to similar messages.
[0006] Messages sent by the message selection system to the users' respective client devices can be opened via an application executing on the respective client device. The application executing on the client device can display messages from the message selection system to the users of the client devices and prompt the users to take an action or elicit a user response. The application can be configured to feed information back to the message selection system including data corresponding to when the message was received at the client device, when the application displayed the message, activities the user performed on the application, activities the user performed on the client device, etc. The message selection system can use the information collected or received from the users' client devices to evaluate the effectiveness of messages in causing users to take an action or the content and timing of user responses to improve the personalization of messages sent to individual users, or in some embodiments, to monitor users and their behavior in response to messages.
[0007] In certain embodiments, the message selection system can maintain user profiles for multiple users. For each user profile, the message selection system can execute an invoker that can invoke a message object at a predetermined time interval. Each message object can generate message candidates using message templates that are candidates for transmission to a client device associated with the user profile. A message object evaluator of the message selection system can incorporate the context data into a model of the message object to output a confidence value associated with the message candidate generated by the message object. The confidence value can indicate the likelihood that the message candidate sent to the user will have an intended effect. The message object evaluator can output a confidence value for the message object that satisfies certain constraints. The message object evaluator can evaluate the confidence value associated with each message candidate based on certain conditions, update the confidence value based on a cooldown factor, and determine to transmit the candidate message to the client device if the updated confidence value exceeds a predetermined value.
[0008] The message selection system can receive response data from a reporting agent of an application running on a client device associated with the user profile. A response evaluator of the message selection system can incorporate the response data into a model of the message object to improve the performance of the model based on the response data corresponding to the previous message. The response data can be compared to a target value (or desired endpoint) of the message object to improve the model of the message object. In this manner, the message selection system can learn from the user's message preferences and responses to improve the selection process of future message candidates as well as improve the quality of the confidence values calculated for future message candidates. In this manner, the message selection system can select fewer messages to send to the user's client device while at the same time increasing the likelihood that the user will engage with or respond to the message and achieve the desired endpoint of the message.
[0009] In accordance with at least one aspect of the present disclosure, a method can include identifying, by one or more processors, a plurality of users and a plurality of message objects. The method can include retrieving, by the one or more processors, context data of each of the plurality of users for the one or more message objects. The method can include generating, by the one or more processors, message candidates based on the one or more message objects and the context data. The method can include updating, by the one or more processors, a confidence value for each message candidate based on a cooldown factor to generate an updated confidence value. The method can include selecting, by the one or more processors, the message candidate based on the updated confidence value. The method can include transmitting, by the one or more processors, the selected message candidate based on the updated confidence value exceeding a predetermined threshold.
[0010] This disclosure describes systems and methods for indirectly updating a model of a message object in a messaging system. The messaging system uses responses to non-rhetorical messages to update a model of a message object associated with the non-rhetorical message. Because rhetorical messages do not elicit a response from users, the messaging system may not be able to update a model of a message object associated with a rhetorical message. In the absence of a response from a user receiving a rhetorical message, no information related to the effectiveness or ineffectiveness of the rhetorical message is provided. Furthermore, without a response from a user, the messaging system may not be able to quantify the benefit that the rhetorical message provides to the user.
[0011] To address some of the challenges associated with the lack of response data from rhetorical messages, the present disclosure is directed to indirect updates to a model of message objects in a message selection system. The message selection system can include a response evaluator that can incorporate response data of a model of message objects associated with non-rhetorical messages into a model of message objects associated with rhetorical messages. In this manner, the message selection system can use the response data of the non-rhetorical messages as a proxy for the response data (or lack thereof) of the rhetorical messages. In this manner, the message selection system can select and send a customized rhetorical message to a user's client device without receiving direct feedback regarding the rhetorical message. The message selection system can also send the rhetorical message to the user's client device at a customized time to prevent message fatigue.
[0012] The message selection system may select messages based on the likelihood that the messages are about to achieve a desired endpoint or goal, while adjusting the likelihood based on the time since the last message was delivered to the client device, to reduce the likelihood that the user experiences message fatigue. The message selection system may selectively send or route messages to the user's client device based on the likelihood that the messages will have the intended effect, thereby reducing the likelihood that the user experiences message fatigue. Because the number and frequency of messages that cause individual users to experience message fatigue may vary, the message selection system may incorporate information about each user's message preferences, application engagement, and responses to similar messages to build a framework or model for delivering customized messages at the appropriate time.
[0013] Messages sent by the message selection system to each of the users' client devices can be opened via an application executing on the respective client device. The application executing on the client device can display messages from the message selection system to a user of the client device and prompt the user to take an action or elicit a user response. The application can be configured to feed information back to the message selection system, including data corresponding to when the message was received at the client device, when the application displayed the message, the activity the user performed on the application, the activity the user performed on the client device, etc. The message selection system can use the information collected or received from the users' client devices to evaluate the effectiveness of messages in persuading users to take an action or the content and timing of user responses, improve personalization of messages sent to individual users, and in some embodiments, monitor users and their activity in response to messages.
[0014] In certain embodiments, the message selection system can maintain user profiles for multiple users. For each user profile, the message selection system can execute an invoker that can invoke a message object at a predetermined time interval. Each message object can generate message candidates using message templates that are candidates for transmission to a client device associated with the user profile. A message object evaluator of the message selection system can incorporate the context data into a model of the message object and output a confidence value associated with the message candidate generated by the message object. The confidence value can indicate the likelihood that the message candidate sent to the user will have an intended effect. The message object evaluator can output a confidence value for the message object that satisfies certain constraints. The message object evaluator can evaluate a confidence value associated with each message candidate based on certain conditions, update the confidence value based on a cooldown factor, and decide to transmit the message candidate to the client device if the updated confidence value exceeds a predetermined threshold.
[0015] The message selection system can receive response data from a reporting agent of an application executing on a client device associated with the user profile. A response evaluator of the message selection system can incorporate the response data into a model of the message object to improve the performance of the model based on the response data corresponding to a previous message sent to the client device. The response data can be compared to a target value (or desired endpoint) of the message object to improve the model of the message object. In this manner, the message selection system can learn from the user's message preferences and responses to improve the selection process of future candidate messages as well as improve the quality of the confidence values calculated for future candidate messages. In doing so, the message selection system can select fewer messages to send to the user's client device while increasing the likelihood that the user will engage with or respond to the message and achieve the desired endpoint of the message.
[0016] The response evaluator can incorporate the response data in the model of the message object associated with the non-rhetorical message into the model of the message object associated with the rhetorical message. In this manner, the message selection system can use the response data of the non-rhetorical message as a surrogate for the response data (or lack thereof) of the rhetorical message. In this manner, the message selection system can select and send an appropriate, personalized rhetorical message to the user's client device even though the message selection system does not receive direct feedback from the user when the user receives the rhetorical message.
[0017] According to at least one aspect of the present disclosure, the method can include identifying, by one or more processors of the server, a plurality of non-rhetorical messages. The non-rhetorical messages can include a first set of requests configured to generate, at the remote device, a response message to be sent to the server. The method can include identifying, by one or more processors of the server, a rhetorical message. The rhetorical messages can include a second set of requests configured to not generate, at the remote device, a response message. The method can include identifying, by one or more processors of the server, a rhetorical message model. The rhetorical message model can correspond to the rhetorical message. The method can include identifying, by one or more processors of the server, a plurality of non-rhetorical message models. The non-rhetorical message model can correspond to the rhetorical message model. The method can include receiving, by one or more processors of the server, a response message to the non-rhetorical message. The method can include updating the rhetorical message model based on the received response message to the non-rhetorical message.
[0018] According to at least one aspect, the present disclosure is directed to a system and method for transmitting a message over a network environment. At least one server having one or more processors may identify a plurality of message object candidates. Each of the plurality of message object candidates may be associated with selection criteria for presenting to a user for achieving one or more of the plurality of endpoints. The at least one server may obtain, from a user activity log associated with the remote computing device, one or more actions by a user recorded via the remote computing device. The at least one server may determine, from the one or more recorded actions by the user, a user state used to identify one of the plurality of message object candidates. The at least one server may determine, for each of the plurality of message object candidates, a confidence value based on a comparison of the user state to the selection criteria of the message object candidate. The confidence value may indicate a prediction of the effectiveness of the corresponding message candidate to the user for achieving one or more of the plurality of endpoints. The at least one server may update, for each of the plurality of message object candidates, the confidence value based at least on at least one elapsed time from a sending time of a previous message corresponding to at least one of the plurality of message object candidates to a current time. The at least one server may select a message object from among the plurality of candidate message objects based on a trust value for the message. The at least one server may generate a message based on the message object selected from the plurality of message objects. The at least one server may transmit the message for presentation to a remote computing device associated with the user.
[0019] In some embodiments, the at least one server may record in an activity log one or more subsequent actions by the user in response to the submission of the message via the remote computing device. In some embodiments, the at least one server may update a user state based on the one or more subsequent actions in response to the submission of the message.
[0020] In some embodiments, the at least one server may establish a message object model using a training data set. The training data set may include historical response data from users to messages corresponding to one or more of the plurality of candidate message objects. In some embodiments, the at least one server may apply the message object model to user states and selection criteria of the plurality of candidate message objects to determine a confidence value.
[0021] In some embodiments, each message object candidate of the multiple message object candidates may include a message template for generating a corresponding message, a model for determining a trust value, a time constraint for sending the corresponding message, and selection criteria defining one or more of the multiple endpoints to achieve within a defined time window.
[0022] In some embodiments, the at least one server may determine a user state corresponding to at least one of the multiple endpoints from a plurality of user states for a user associated with the remote computing device. In some embodiments, the at least one server may identify and apply to the trust value a cool-down factor corresponding to an elapsed time between a last transmission time and a current time.
[0023] In some embodiments, the at least one server may select a message object from the plurality of candidate message objects based on a comparison of the trust value for the message object to a threshold. The threshold may be defined by a delivery policy for sending the message. In some embodiments, the at least one server may obtain, from a user activity log, one or more actions by a user in response to a previous message submission corresponding to one or more of the plurality of candidate message objects.
[0024] In some embodiments, the at least one server may send the message to the remote computing device at a current time defined by a delivery policy for sending the message. In some embodiments, the at least one server may remove at least a subset of the plurality of message object candidates from the selection based on a user state not matching a selection criterion for each subset.
[0025] According to at least one aspect, the present disclosure contemplates a system and method for selecting a message for transmission over a network environment. At least one server having one or more processors may identify a plurality of message objects. Each message object of the plurality of message objects may be associated with one or more of a plurality of endpoints to be achieved. The at least one server may determine, for each message object of the plurality of message objects, a number of users associated with a plurality of remote computing devices to which a corresponding message has been submitted that satisfy at least one of the one or more of a plurality of endpoints of the message object. The at least one server may determine, for each message object of the plurality of message objects, a performance score for the message object based on a number of users to which a corresponding message has been submitted that satisfy at least one of the one or more of a plurality of action endpoints of the message object. The at least one server may rank the plurality of message objects based on the corresponding performance scores. The at least one server may restrict at least a subset of the plurality of message objects from being used to generate a message for transmission to the plurality of remote computing devices based on the ranking of the plurality of message objects.
[0026] In some embodiments, the at least one server may identify a plurality of message objects of a first message type. Each message object of the plurality of message objects of the first message type may define a message template for generating a corresponding message. The corresponding message may have one or more input elements when presented on the plurality of remote computing devices.
[0027] In some embodiments, at least the server may identify a second plurality of message objects, each message object of the second plurality of message objects associated with one or more endpoints to be achieved. In some embodiments, at least the server may select a subset of the second plurality of message objects for transmission based on one or more of the plurality of endpoints to be achieved associated with a remaining subset of the plurality of message objects of the first message type following removal of at least the subset.
[0028] In some embodiments, each message object of the second plurality of message objects may be of a second message type different from the first message type of the plurality of message objects. Each message object of the second plurality of message objects of the second message type may define a message template for generating a corresponding message. The corresponding message may lack an input element when presented on the multiple remote computing devices.
[0029] In some embodiments, for each message object of the plurality of message objects, the at least one server may identify, in each response to a message corresponding to the message object when presented to one of the plurality of remote computing devices, inputs for one or more input elements of the corresponding message. In some embodiments, for each message object of the plurality of message objects, the at least one server may determine, in each response to the corresponding message, that at least one of a plurality of endpoints of the message object has been satisfied based on inputs on one or more input elements of the message. In some embodiments, the at least one server may determine, for each message object of the plurality of messages, a number of users based on a determination that at least one of a plurality of endpoints of the message object has been satisfied.
[0030] In some embodiments, the at least one server may establish a performance model for determining a performance score using a training data set. The training data set may include historical response data from users to one or more of the plurality of candidate message objects. In some embodiments, the at least one server may apply the performance model to responses received from the plurality of remote computing devices in response to submission of the corresponding messages.
[0031] In some embodiments, the at least one server may determine, for each message object of the plurality of message objects, that at least one of the plurality of endpoints of the message object is satisfied based on input to one or more input elements of a message corresponding to the message object. In some embodiments, the at least one server may update a performance model for determining a performance score with a determination, for each message object of the plurality of message objects, that at least one of the plurality of endpoints of the message object is satisfied.
[0032] In some embodiments, the at least one server may attribute one or more of the multiple endpoints to each message object of the multiple message objects based on responses received from the multiple remote computing devices in response to submission of the corresponding message. In some embodiments, the at least one server may restrict at least a subset of the multiple message objects based on a comparison of the performance score to a threshold. In some embodiments, the at least one server may permit transmission of a plurality of messages corresponding to a remaining subset of the plurality of message objects between the multiple remote computing devices.
[0033] The foregoing and other objects, aspects, features, and advantages of the present disclosure will become more apparent and be better understood by referring to the following description taken in conjunction with the accompanying drawings. [Brief description of the drawings]
[0034] [Figure 1A] FIG. 1 is a block diagram illustrating an embodiment of a network environment including a local device in communication with a remote device. [Figure 1B] FIG. 1 is a block diagram illustrating one embodiment of a computer useful in connection with the methods and systems described herein. [Figure 1C] FIG. 1 is a block diagram illustrating one embodiment of a computer useful in connection with the methods and systems described herein. [Figure 1D] FIG. 1 is a block diagram illustrating one embodiment of a computer useful in connection with the methods and systems described herein. [Diagram 2] FIG. 1 is a block diagram illustrating one embodiment of a system for sending relevant messages to a user. [Diagram 3] FIG. 11 is a flow diagram illustrating an example method for contention message selection. [Figure 4] 1 is a graph of confidence values versus time with sharp peaks. [Diagram 5] 13 is a graph of confidence versus time using upper and lower thresholds. [Figure 6A] This shows the distribution of the number of messages sent per day. [Figure 6B] Shows the distribution of the number of messages sent per day [Figure 6C] Shows the distribution of the number of messages sent per day [Figure 7] FIG. 1 is a block diagram illustrating one embodiment of a system for sending relevant messages to a user. [Figure 8] FIG. 1 is a flow diagram illustrating an example of a method for updating an indirect learning model. [Figure 9A]1 is a block diagram of a system for selecting and sending messages in a network environment using trust values. [Figure 9B] FIG. 1 is a sequence diagram of a system for selecting and sending messages in a network environment using trust values. [Figure 9C] FIG. 1 is a sequence diagram of a system for selecting and sending messages in a network environment using trust values. [Figure 9D] FIG. 1 is a sequence diagram of a system for selecting and sending messages in a network environment using trust values. [Figure 10] 1 is a flow diagram of a method for selecting and sending messages in a network environment using trust values. [Figure 11A] 1 is a block diagram of a system for selecting and sending different types of messages in a network environment. [Figure 11B] FIG. 1 is a sequence diagram of a system for selecting and sending different types of messages in a network environment. [Figure 11C] FIG. 1 is a sequence diagram of a system for selecting and sending different types of messages in a network environment. [Figure 11D] FIG. 1 is a sequence diagram of a system for selecting and sending different types of messages in a network environment. [Figure 12] 1 is a flow diagram of a method for selecting and sending different types of messages in a networked environment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0035] The features and advantages of the present invention will become more apparent from the following detailed description when taken in conjunction with the drawings, in which like reference characters indicate corresponding elements throughout. In the drawings, like reference numbers generally indicate identical, functionally similar, or structurally similar elements.
[0036] For purposes of reading the following description of the various embodiments, it may be helpful to list the sections of this specification and their respective contents as follows:
[0037] Section A describes network and computing environments useful for practicing the embodiments described herein.
[0038] Section B describes embodiments of systems and methods for selecting and sending messages based on trust values and cool-down factors in a network environment.
[0039] Section C describes embodiments of systems and methods for indirect updates to a messaging system.
[0040] Section D describes embodiments of systems and methods that use trust values to select and send messages across a network environment.
[0041] Section E describes embodiments of systems and methods for selecting and sending different types of messages in a network environment.
[0042] A. Network and Computing Environment
[0043] In addition to describing specific embodiments of the present solution, it is useful to describe aspects of the operating environment and associated system components (e.g., hardware elements) in connection with the methods and systems described herein. Referring to FIG. 1A, one embodiment of a network environment is depicted. In brief overview, the network environment includes one or more clients 102a-102n (also commonly referred to as local machines(s) 102, clients 102, client nodes 102, client machines 102, client computers 102, client devices 102, endpoints 102, or endpoint nodes 102) in communication with one or more servers 106a-106n (also commonly referred to as servers 106, nodes 106, or remote machines 106) over one or more networks 104. In some embodiments, a client 102 has the capability to function both as a client node seeking access to resources provided by a server, and as a server providing access to hosted resources for other clients 102a-102n.
[0044] Although FIG. 1A shows a network 104 between the client 102 and the server 106, the client 102 and the server 106 may be on the same network 104. In some embodiments, there are multiple networks 104 between the client 102 and the server 106. In one of these embodiments, the network 104' (not shown) may be a private network and the network 104 may be a public network. In another of these embodiments, the network 104 may be a private network and the network 104' may be a public network. In yet another of these embodiments, the networks 104 and 104' may both be private networks. Referring to FIG. 1A, one embodiment of a network environment is depicted. In brief overview, the network environment includes one or more clients 102a-102n (also commonly referred to as local machines(s) 102, clients 102, client nodes 102, client machines 102, client computers 102, client devices 102, endpoints 102, or endpoint nodes 102) in communication with one or more servers 106a-106n (also commonly referred to as servers 106, nodes 106, or remote machines 106) over one or more networks 104. In some embodiments, the clients 102 have the capability to function both as client nodes seeking access to resources provided by a server and as servers providing access to hosted resources for other clients 102a-102n.
[0045] The networks 104 may be connected via wired or wireless links. Wired links may include digital subscriber lines (DSL), coaxial cable lines, or fiber optic lines. Wireless links may include BLUETOOTH, Wi-Fi, Worldwide Interoperability for Microwave Access (WiMAX), infrared channels, or satellite bands. Wireless links may also include cellular network standards used for communication between mobile devices, including standards that fall under any of 1G, 2G, 3G, and 4G. Network standards may qualify as one or more generations of mobile communication standards by meeting standards such as those established by the International Telecommunications Union. For example, the 3G standard may correspond to the International Mobile Telecommunications-2000 (IMT-2000) specifications, and the 1G standard may correspond to the International Mobile Telecommunications-Advanced (IMT-Advanced) specifications. Examples of cellular network standards include AMPS, GSM, GPRS, UMTS, LTE, LTE Advanced, Mobile WiMAX, and WiMAX Advanced. Cellular network standards may use a variety of channel access schemes, such as FDMA, TDMA, CDMA, SDMA, etc. In some embodiments, different types of data may be transmitted over different links and standards. In other embodiments, the same type of data may be transmitted over different links and standards.
[0046] The network 104 may be of any type or form of network. The geographical scope of the network 104 may vary widely, and the network 104 may be a body area network (BAN), a personal area network (PAN), a local area network (LAN), such as an intranet, a metropolitan area network (MAN), a wide area network (WAN), or the Internet. The topology of the network 104 may be of any form, including, for example, any of point-to-point, bus, star, ring, mesh, or tree. The network 104 may be virtual, an overlay network that sits on top of one or more layers of other networks 104'. The network 104 may be any such network topology known to those skilled in the art that can support the operations described herein. The network 104 may utilize layers or stacks of different technologies and protocols, including, for example, Ethernet protocol, Internet Protocol Suite (TCP / IP), ATM (Asynchronous Transfer Mode) technology, SONET (Synchronous Optical Network) protocol, or SDH (Synchronous Digital Hierarchy) protocol. The TCP / IP Internet protocol suite may include an application layer, a transport layer, an Internet layer (including, for example, IPv6), or a link layer. Network 104 may be a broadcast network, a telecommunications network, a data communications network, or a type of computer network.
[0047] In some embodiments, the system may include multiple logically grouped servers 106. In one of these embodiments, the logical grouping of servers may be referred to as a server farm 38 or a machine farm 38. In another of these embodiments, the servers 106 may be geographically distributed. In other embodiments, the machine farm 38 may be managed as a single entity. In yet other embodiments, the machine farm 38 includes multiple machine farms 38. The servers 106 within each machine farm 38 may be heterogeneous, with one or more of the servers 106 or machines 106 being capable of operating according to one type of operating system platform (e.g., WINDOWS NT manufactured by Microsoft Corporation of Redmond, Washington), while one or more of the other servers 106 are capable of operating according to another type of operating system platform (e.g., Unix, Linux, or Mac OS X).
[0048] In one embodiment, the servers 106 in the machine farm 38, along with their associated storage systems, may be housed in a high density rack system and located in an enterprise data center. In this embodiment, consolidating the servers 106 in this manner allows the servers 106 and high performance storage systems to be located on a localized, high performance network, improving system manageability, data security, system physical security, and system performance. The centralization of the servers 106 and storage systems, combined with advanced system management tools, allows for more efficient use of server resources.
[0049] The servers 106 of each machine farm 38 need not be physically close to another server 106 of the same machine farm 38. Thus, a group of servers 106 logically grouped as a machine farm 38 may be interconnected using wide area network (WAN) or metropolitan area network (MAN) connections. For example, a machine farm 38 may include servers 106 physically located in different continents or different regions, such as a country, state, city, campus, room, etc. Data transmission rates between the servers 106 in a machine farm 38 may be increased if the servers 106 are connected using local area network (LAN) connections or some form of direct connection. Furthermore, a heterogeneous machine farm 38 may include one or more servers 106 that operate according to one type of operating system, while one or more other servers 106 may run one or more types of hypervisors rather than an operating system. In these embodiments, a hypervisor may be used to emulate virtual hardware, partition physical hardware, virtualize physical hardware, and run virtual machines that provide access to a computing environment, allowing multiple operating systems to run simultaneously on a host computer. Native hypervisors can run directly on a host computer. Examples of hypervisors include VMware ESX / ESXi from VMWare, headquartered in Palo Alto, California, the open source Xen hypervisor developed by Citrix Systems, and the HYPER-V hypervisor provided by Microsoft and other companies. Hosted hypervisors run at the second software level within the operating system. Examples of hosted hypervisors include VMware Workstation and VIRTUALBOX.
[0050] Management of the machine farm 38 may be distributed. For example, one or more servers 106 may comprise components, subsystems, and modules that support one or more management services for the machine farm 38. In one of these embodiments, the one or more servers 106 provide dynamic data management capabilities, including failover, data replication, and processing techniques to increase the robustness of the machine farm 38. Each server 106 may communicate with persistent storage, and in some embodiments, with dynamic storage. The servers 106 may be file servers, application servers, web servers, proxy servers, appliances, network appliances, gateways, gateway servers, virtualization servers, deployment servers, SSL VPN servers, or firewalls. In one embodiment, the servers 106 may also be referred to as remote machines or nodes.
[0051] Referring to FIG. 1B, a cloud computing environment is depicted. The cloud computing environment may provide one or more resources provided by a network environment to a client 102. The cloud computing environment may include one or more clients 102a-102n that communicate with a cloud 108 via one or more networks 104. The clients 102 may include, for example, thick clients, thin clients, and zero clients. A thick client may provide at least some functionality even when disconnected from the cloud 108 or server 106. A thin client or zero client may rely on a connection to the cloud 108 or server 106 to provide functionality. A zero client may rely on the cloud 108 or other networks 104 or server 106 to obtain operating system data for the client device. The cloud 108 may include a back-end platform, for example, a server 106, storage, a server farm, or a data center.
[0052] The cloud 108 may be public, private, or hybrid. A public cloud may include a public server 106 maintained by a third party to the client 102 or the client's owner. The server 106 may be located off-site at a remote geographic location as disclosed above or otherwise. A public cloud may be connected to the server 106 via a public network. A private cloud may include a private server 106 physically maintained by the client 102 or the client's owner. A private cloud may be connected to the server 106 via a private network 104. A hybrid cloud 108 may include both a private network 104 and a public network 106.
[0053] Cloud 108 may also include cloud-based delivery, such as Software as a Service (SaaS) 110, Platform as a Service (PaaS) 112, and Infrastructure as a Service (IaaS) 114. IaaS may refer to a user renting the use of infrastructure resources needed for a specific period of time. IaaS providers may provide storage, network, server, or virtualization resources from a large pool, allowing users to access more resources as needed and scale up quickly. Examples of IaaS include AMAZON WEB SERVICES from Amazon, Inc., Seattle, Washington; RACKSPACE CLOUD from Rackspace US, Inc., San Antonio, Texas; Google Compute Engine from Google, Inc., Mountain View, California; and RIGHTSCALE from RightScale, Inc., Santa Barbara, California. PaaS providers may provide additional resources such as operating systems, middleware, and runtime resources in addition to the features offered by IaaS, such as storage, network, server, and virtualization. Examples of PaaS include WINDOWS AZURE from Microsoft Corporation of Redmond, Washington, Google App Engine from Google, Inc., and HEROKU from Heroku, Inc. of San Francisco, California. A SaaS provider may provide the resources that the PaaS provides, such as storage, network, server, virtualization, operating system, middleware, or runtime resources. In some embodiments, the SaaS provider may provide additional resources including, for example, data and application resources.Examples of SaaS include GOOGLE APPS provided by Google, SALESFORCE provided by Salesforce, Inc. of San Francisco, California, or OFFICE 365 provided by Microsoft. Examples of SaaS may also include data storage providers such as DROPBOX provided by Dropbox, Inc. of San Francisco, California, Microsoft SKYDRIVE provided by Microsoft, Google Drive provided by Google, or Apple ICLOUD provided by Apple Inc. of Cupertino, California.
[0054] Client 102 may access IaaS resources using one or more IaaS standards, including, for example, Amazon Elastic Compute Cloud (EC2), Open Cloud Computing Interface (OCCI), Cloud Infrastructure Management Interface (CIMI), or OpenStack standards. Some IaaS standards may allow clients to access resources over HTTP, may use the Representational State Transfer (REST) protocol, or the Simple Object Access Protocol (SOAP). Client 102 may access PaaS resources using different PaaS interfaces. Some PaaS interfaces use HTTP packages, standard Java APIs, JavaMail APIs, Java Data Objects (JDO), Java Persistence API (JPA), web integration APIs for different programming languages, including Python APIs, such as Rack for Ruby, WSGI for Python, or PSGI for Perl, or other APIs that may be built on REST, HTTP, XML, or other protocols. Client 102 may access SaaS resources using a web-based user interface provided by a web browser (e.g., GOOGLE CHROME, Microsoft Internet Explorer, or Mozilla Firefox from Mozilla, Inc. of Mountain View, Calif.). Client 102 may also access SaaS resources through a smartphone or tablet application, including, for example, the Salesforce Sales Cloud or Google Drive app. Client 102 may also access SaaS resources through a client's operating system, including, for example, the Windows file system from Dropbox, Inc.
[0055] In some embodiments, access to IaaS, PaaS, or SaaS resources may be authenticated. For example, a server or authentication server may authenticate a user via a security certificate, HTTPS, or an API key. The API key may include various encryption standards, such as Advanced Encryption Standard (AES). Data resources may be transmitted over Transport Layer Security (TLS) or Secure Socket Layer (SSL).
[0056] The client 102 and the server 106 may be deployed or implemented as any type and form of computing device, such as a computer, network device, or facility, capable of communicating over any type and form of network and performing the operations described herein. FIGS. 1C and 1D show block diagrams of a computing device 100 useful for implementing an embodiment of the client 102 or the server 106. As shown in FIGS. 1C and 1D, each computing device 100 includes a central processing unit 121 and a main memory unit 122. As shown in FIG. 1C, the computing device 100 may include a storage device 128, an installation device 116, a network interface 118, an I / O controller 123, display devices 124a-124n, a keyboard 126, and a pointing device 127 (e.g., a mouse). The storage device 128 may include, but is not limited to, an operating system, software, and software for a message selection system (MSS) 197. As shown in FIG. 1D, each computing device 100 may also include additional optional elements, such as a memory port 103, a bridge 170, one or more I / O devices 130a-130n (collectively referred to using the reference numeral 130), and a cache memory 140 in communication with the central processing unit 121.
[0057] Central processing unit 121 is any logic circuit that responds to and processes instructions fetched from main memory device 122. In many embodiments, central processing unit 121 is provided by a microprocessor unit, such as those manufactured by Intel Corporation of Mountain View, Calif., Motorola Corporation of Schaumburg, Ill., ARM processors and TEGRA systems on a chip (SoCs) manufactured by Nvidia, Inc. of Santa Clara, Calif., POWER7 processors manufactured by International Business Machines (IBM) of White Plains, New York, or Advanced Micro Devices, Inc. of Sunnyvale, Calif. Computing device 100 may be based on any of these processors or other processors capable of operating as described herein. Central processing unit 121 may utilize instruction level parallelism, thread level parallelism, different levels of caches, and multi-core processors. Multi-core processors may include two or more processing units on a single computing component. Examples of multi-core processors include the AMD PHENOM IIX2, the INTEL CORE i5, and the INTEL CORE i7.
[0058] The main memory device 122 may include one or more memory chips that can store data and allow any memory location to be directly accessed by the microprocessor 121. The main memory device 122 may be volatile and faster than the storage 128 memory. The main memory device 122 may be dynamic random access memory (DRAM) or any variant including static random access memory (SRAM), burst SRAM or synchronous burst SRAM (BSRAM), fast page mode DRAM (FPM DRAM), enhanced DRAM (EDRAM), extended data output RAM (EDO RAM), extended data output DRAM (EDO DRAM), burst extended data output DRAM (BEDO DRAM), single data rate synchronous DRAM (SDR SDRAM), double data rate SDRAM (DDR SDRAM), direct Rambus DRAM (DRDRAM) or extreme data rate DRAM (XDR DRAM). In some embodiments, the main memory 122 or storage 128 may be non-volatile, such as non-volatile read access memory (NVRAM), flash memory non-volatile static RAM (nvSRAM), ferroelectric RAM (FeRAM), magnetoresistive RAM (MRAM), phase change memory (PRAM), conductive bridge RAM (CBRAM), silicon-oxide-nitride-oxide-silicon (SONOS), resistive RAM (RRAM), Racetrack, nanoRAM (NRAM), or millipede memory. The main memory 122 may be based on any of the memory chips mentioned above or other available memory chips operable as described herein. In the embodiment shown in FIG. 1C, the processor 121 communicates with the main memory 122 via a system bus 150 (described in more detail below). FIG. 1D illustrates an embodiment of the computing device 100 in which the processor communicates directly with the main memory 122 via memory port 103. For example, in FIG. 1D, the main memory 122 may be a DRDRAM.
[0059] 1D illustrates an embodiment in which the main processor 121 communicates directly with the cache memory 140 via a secondary bus, sometimes referred to as a backside bus. In other embodiments, the main processor 121 communicates with the cache memory 140 using a system bus 150. The cache memory 140 typically has a faster response time than the main memory 122 and is typically provided by SRAM, BSRAM, or EDRAM. In the embodiment illustrated in FIG. 1D, the processor 121 communicates with various I / O devices 130 via a local system bus 150. Various buses can be used to connect the central processing unit 121 to any of the I / O devices 130, such as a PCI bus, a PCI-X bus, or a PCI-Express bus, or a NuBus. In an embodiment in which the I / O device is a video display 124, the processor 121 may communicate with the display 124 or an I / O controller 123 for the display 124 using an Advanced Graphics Port (AGP). 1D illustrates an embodiment of computer 100 in which main processor 121 communicates directly with I / O device 130b or other processors 121' via HYPERTRANSPORT, RAPIDIO, or INFINIBAND communications technologies. Also illustrated in FIG. 1D is an embodiment of a mix of local bus and direct communications, in which processor 121 communicates directly with I / O device 130b while using a local interconnect bus to communicate with I / O device 130a.
[0060] A wide variety of I / O devices 130a-130n may be present in computing device 100. Input devices may include a keyboard, mouse, track pad, track ball, touch pad, touch mouse, multi-touch pad and touch mouse, microphone, multi-array microphone, drawing tablet, camera, single lens reflex camera (SLR), digital single lens reflex camera (DSLR), CMOS sensor, acceleration sensor, infrared light sensor, pressure sensor, magnetic sensor, angular velocity sensor, depth sensor, proximity sensor, ambient light sensor, gyro sensor, or other sensor. Output devices may include a video display, a graphics display, speakers, headphones, an inkjet printer, a laser printer, a 3D printer, and the like.
[0061] Devices 130a-130n may include a combination of input or output devices, including, for example, Microsoft KINECT, a Wiimote for Nintendo Wii, a Gamepad for Nintendo Wii U, or an Apple iPhone. Some devices 130a-130n allow for gesture recognition input by combining some of the inputs and outputs. Some devices 130a-130n provide facial recognition that can be used as input for different purposes, including authentication and other commands. Some devices 130a-130n provide voice recognition and input, including, for example, Microsoft KINECT, SIRI for iPhone by Apple, Google Now, or Google Voice Search.
[0062] Additional devices 130a-130n may have both input and output capabilities, such as, for example, haptic feedback devices, touch screen displays, or multi-touch displays. Touch screens, multi-touch displays, touch pads, touch mice, or other touch sensing devices may use different technologies for sensing touch, including, for example, capacitive, surface capacitive, projected capacitive touch (PCT), in-cell capacitive, resistive, infrared, waveguide, distributed signal touch (DST), in-cell optical, surface acoustic wave (SAW), bending wave touch (BWT), or force-based sensing technologies. Some multi-touch devices may provide more than one contact point on the surface to enable advanced functionality, including gestures such as pinch, spread, rotate, scroll, etc. Some touch screen devices, including, for example, the Microsoft PIXELSENSE or the Multi-Touch Collaboration Wall, may have larger surfaces, such as table tops or wall tops, and may still interact with other electronic devices. Some of the I / O devices 130a-130n, display devices 124a-124n, or device groups may be augmented reality devices. The I / O devices may be controlled by an I / O controller 123, as shown in FIG. 1C. The I / O controller may control one or more I / O devices, such as, for example, a keyboard 126 and a pointing device 127, such as, for example, a mouse or an optical pen. Additionally, the I / O devices may provide a storage or installation device 116 for the computing device 100. In yet another embodiment, the computing device 100 may provide a USB connection (not shown) for receiving a handheld USB storage device. In a further embodiment, the I / O device 130 may be a bridge between the system bus 150 and an external communication bus (e.g., a USB bus, a SCSI bus, a FireWire bus, an Ethernet bus, a Gigabit Ethernet bus, a Fibre Channel bus, or a Thunderbolt bus).
[0063] In some embodiments, display devices 124a-124n may be connected to I / O controller 123. The display devices may include, for example, a liquid crystal display (LCD), a thin film transistor LCD (TFT-LCD), a blue-phase LCD, an electronic paper (e-ink) display, a flexile display, a light emitting diode display (LED), a digital light processing (DLP) display, a liquid crystal on silicon (LCOS) display, an organic light emitting diode (OLED) display, an active matrix organic light emitting diode (AMOLED) display, a liquid crystal laser display, a time multiplexed optical shutter (TMOS) display, or a 3D display. Examples of 3D displays may use, for example, stereoscopy, polarizing filters, active shutters, or autostereoscopy. Also, display devices 124a-124n may be head mounted displays (HMDs). In some embodiments, display devices 124a-124n or corresponding I / O controller 123 may be controlled via an OPENGL or DIRECTX API or other graphics library, or may have hardware support.
[0064] In some embodiments, computing device 100 may include or be connected to multiple display devices 124a-124n, each of which may be of the same type or different types or forms. As such, any of I / O devices 130a-130n or I / O controller 123 may include any type or form of suitable hardware, software, or combination of hardware and software to support, enable, or provide for the connection and use of multiple display devices 124a-124n by computing device 100. For example, computing device 100 may include any type or form of video adapter, video card, driver, or library to interface, communicate, connect, or otherwise use display devices 124a-124n. In one embodiment, a video adapter may include multiple connectors for interfacing to multiple display devices 124a-124n. In other embodiments, computing device 100 may include multiple video adapters, each video adapter connected to one or more display devices 124a-124n. In some embodiments, any portion of the operating system of the computing device 100 may be configured to use multiple displays 124a-124n. In other embodiments, one or more of the display devices 124a-124n may be provided by one or more other computing devices 100a or 100b connected to the computing device 100 via the network 104. In some embodiments, software may be designed and constructed to use the display device of another computer as a second display device 124a for the computing device 100. For example, in one embodiment, an Apple iPad may be connected to the computing device 100 and use the display of the computing device 100 as an additional display screen that can be used as an extended desktop.Those skilled in the art will recognize and understand the various ways and embodiments in which a computing device 100 may be configured to have multiple display devices 124a-124n.
[0065] Referring again to FIG. 1C, the computing device 100 may include storage 128 (e.g., one or more hard disk drives or a redundant array of independent disks) for storing an operating system or other related software, and for storing application software programs, such as any programs related to the delivery scheduler 198 software. Examples of storage 128 include, for example, hard disk drives (HDD); optical drives, including CD drives, DVD drives, or BLU-RAY drives; solid state drives (SSD); USB flash drives; or any other device suitable for storing data. Some storage devices may include multiple volatile and non-volatile memories, including, for example, solid state hybrid drives that combine a hard disk and a solid state cache. Some storage devices 128 may be non-volatile, changeable, or read-only. Some storage devices 128 may be internal and connected to the computing device 100 via a bus 150. Some storage devices 128 may be external and connected to the computing device 100 via an I / O device 130 that provides an external bus. Several storage devices 128 may be connected to the computing device 100 via the network interface 118 over the network 104, including, for example, a remote disk for the MACBOOK AIR by Apple. Some client devices 100 may not require non-volatile storage 128 and may be thin clients or zero clients 102. Also, some storage devices 128 may be used as installation devices 116 and suitable for installing software and programs. Additionally, the operating system and software may be executed from a bootable medium, for example, a bootable CD, for example, KNOPPIX, a bootable CD for GNU / Linux available as a GNU / Linux distribution from knoppix.net.
[0066] The client device 100 may also install software or applications from an application distribution platform. Examples of application distribution platforms include the App Store for iOS provided by Apple, the Mac App Store provided by Apple, GOOGLE PLAY for Android OS provided by Google, Chrome Webstore for CHROME OS provided by Google, and Amazon Appstore for Android OS and KINDLE FIRE provided by Amazon. The application distribution platform may facilitate the installation of software on the client device 102. The application distribution platform may include a repository of applications on the server 106 or cloud 108 that the clients 102a-102n access via the network 104. The application distribution platform may include applications developed and provided by various developers. A user of the client device 102 may select, purchase, or download an application via the application distribution platform.
[0067] Additionally, the computing device 100 may be connected to any of a variety of networks, including, but not limited to, standard telephone line LAN or WAN links (e.g., 802.1, T1, T3, Gigabit Ethernet, Infiniband), broadband connections (e.g., ISDN, Frame Relay, ATM, Gigabit Ethernet, Ethernet-over-SONET, ADSL, VDSL, BPON, GPON, fiber optic including FiOS), wireless connections, or a combination of any or all of the above. Connections may be established using a variety of communication protocols (e.g., TCP / IP, Ethernet, ARCNET, SONET, SDH, Fiber Distributed Data Interface (FDDI), IEEE 802.1a / b / g / n / ac CDMA, GSM, WiMax, direct asynchronous connections). In one embodiment, computing device 100 communicates with other computing devices 100' via any type or form of gateway or tunneling protocol, such as Secure Sockets Layer (SSL) or Transport Layer Security (TLS), or the Citrix Gateway Protocol manufactured by Citrix Systems, Inc. of Fort Lauderdale, Fla. Network interface 118 may comprise an internal network adapter, a network interface card, a PCMCIA network card, an EXPRESSCARD network card, a cardbus network adapter, a wireless network adapter, a USB network adapter, a modem, or other device suitable for interfacing with any type of network capable of communicating with computing device 100 and performing the operations described herein.
[0068] 1B and 1C may operate under the control of an operating system that controls the scheduling of tasks and access to system resources. Computing device 100 may run any operating system, such as any version of the Microsoft WINDOWS operating system, different releases of Unix and Linux operating systems, any version of MAC OS for Macintosh computers, any embedded operating system, any real-time operating system, any open source operating system, any special-purpose operating system, any operating system for mobile computing devices, or any other operating system capable of running on a computing device and performing the operations described herein. Representative operating systems include, but are not limited to, WINDOWS 2000, WINDOWS Server 2012, WINDOWS CE, WINDOWS Phone, WINDOWS XP, WINDOWS VISTA, and WINDOWS 7, WINDOWS RT, WINDOWS 8 manufactured by Microsoft Corporation of Redmond, Washington, MAC OS and iOS manufactured by Apple Inc. of Cupertino, California, as well as freely available operating systems such as Linux, such as the Linux Mint distribution ("Distro") or Ubuntu distributed by Canonical, Ltd. of London, England, or Unix and other Unix-like derivatives, and Android developed by Google of Mountain View, California. Some operating systems, including, for example, CHROME OS by Google, may be used on zero or thin clients, including, for example, CHROMEBOOKS.
[0069] The computer system 100 may be a workstation, a telephone, a desktop computer, a laptop or notebook computer, a netbook, an ULTRABOOK, a tablet, a server, a handheld computer, a mobile phone, a smart phone or other portable telecommunication device, a media playback device, a gaming system, a mobile computing device, or any other type or form of computing, telecommunication, or media device capable of communication. The computing system 100 has sufficient processor power and memory capacity to perform the operations described herein. In some embodiments, the computing device 100 may have different processors, operating systems, and input devices consistent with the device. The Samsung GALAXY smartphone, for example, operates under the control of the Android operating system developed by Google. The GALAXY smartphone receives input via a touch interface.
[0070] In some embodiments, computing device 100 is a gaming system. For example, computer system 100 may comprise a PLAYSTATION 3, or a PERSONAL PLAYSTATION PORTABLE (PSP), or a PLAYSTATION VITA device manufactured by Sony Corporation of Tokyo, Japan, a Nintendo DS, Nintendo 3DS, Nintendo WII, or Nintendo WII U device manufactured by Nintendo Co., Ltd. of Kyoto, Japan, an XBOX 360 device manufactured by Microsoft Corporation of Redmond, Washington, USA, or the like.
[0071] In some embodiments, computing device 100 is a digital audio player, such as the Apple IPOD, IPOD Touch, and IPOD NANO lines of devices manufactured by Apple Computer, Inc. of Cupertino, Calif. Some digital audio players may have other functionality, including, for example, gaming systems, or any functionality available through applications from digital application distribution platforms. For example, an IPOD Touch may access the Apple App Store. In some embodiments, computing device 100 is a portable media player or digital audio player that supports file formats, including, but not limited to, MP3, WAV, M4A / AAC, WMA protected AAC, AIFF, Audible audiobook, Apple Lossless audio file formats, and .mov, m4v, and .mp4 MPEG-4 (H.264 / MPEG-4 AVC) video file formats.
[0072] In some embodiments, computing device 100 is a tablet, such as the IPAD line of devices by Apple, Inc., the GALAXY TAB family of devices by Samsung, Inc., or the KINDLE FIRE by Amazon, Inc. of Seattle, Washington. In other embodiments, computing device 100 is an e-book reader, such as the KINDLE family of devices by Amazon, Inc., or the NOOK family of devices by Barnes & Noble, Inc. of New York, New York.
[0073] In some embodiments, the communication device 102 includes a combination of devices, such as a smartphone in combination with a digital audio player or a portable media player. For example, one of these embodiments is a smartphone, such as the Apple IPHONE family of smartphones, the Samsung GALAXY family of smartphones, or the Motorola DROID family of smartphones. In yet another embodiment, the communication device 102 is a laptop or desktop computer equipped with a web browser and a microphone and speaker system, such as a telephony headset. In these embodiments, the communication device 102 is web-enabled and can receive and initiate phone calls. In some embodiments, the laptop or desktop computer also includes a webcam or other video capture device to enable video chats and video calls.
[0074] In some embodiments, the status of one or more machines 102, 106 in the network 104 is monitored, typically as part of network management. In one of these embodiments, the status of a machine may include identifying load information (e.g., the number of processes on the machine, CPU and memory utilization), identifying port information (e.g., the number of available communication ports and port addresses), or identifying session status (e.g., process duration and type, and whether the process is active or idle). In another of these embodiments, this information may be identified by a number of metrics that may be applied, at least in part, toward decisions in load balancing, network traffic management, and network failure recovery, as well as any aspect of the operation of the present solutions described herein. The above-mentioned aspects of the operating environment and components will become apparent in the context of the systems and methods disclosed herein.
[0075] B. Message selection and transmission based on trust value and cooldown factor in a network environment
[0076] Users of personalized messaging systems can suffer from message fatigue, which reduces the effectiveness of messages to their intended recipients. Message fatigue leads to wasted computing resources and bandwidth because messages sent over the network to a user's client device do not serve a purpose on the client device. In applications that require user interaction and response, personalized messaging can be a tool to achieve user engagement goals.
[0077] This disclosure describes systems and methods for selecting and sending messages based on a trust value and a cool-down factor. The personalized messaging system can send relevant messages to users at relevant times. The personalized messaging system can increase user engagement and application effectiveness. Appropriate content delivered at the right time for each user can positively impact the user's engagement with the application. Personalized and relevant messages delivered to each user at the right time through the application can lead to real-world actions taken by the user to improve their health, well-being, fitness, and overall quality of life.
[0078] To address some of the challenges associated with message fatigue, as well as other related challenges, the present disclosure relates to a message selection system that can invoke multiple message objects. These message objects, when invoked, can generate multiple messages. The message selection system can generate a confidence value for each message and use the confidence value associated with each message to select one or more messages to send to an application on a client device. The message selection system can adjust the confidence value associated with each selected message based on a cool-down factor. The message selection system can use the cool-down factor to adjust the confidence value of each of the selected messages based on a variety of factors, including, but not limited to, the amount of time since the last message was received by the client device, the total number of messages that a user of the client device has selected to receive within a given time period, and the amount of time that a user of the client device has been awake or active. The message selection system can then make a decision to send one or more of the selected messages based on the adjusted confidence value of each of the selected messages.
[0079] The message selection system can select messages based on the likelihood that the messages will achieve a desired endpoint or goal based on the time since the last message was delivered to the client device, adjusting the likelihood based on the time since the last message was delivered to the client device to reduce the likelihood that the user will experience message fatigue. The message selection system can selectively send or route messages to the user's client device based on the likelihood that the messages will have the intended effect, thereby reducing the likelihood that the user will experience message fatigue. Because the number and frequency of messages that cause individual users to experience message fatigue may vary, the message selection system can incorporate information about each user's message preferences, application engagement, and reactions to similar messages to build a framework or model for delivering personalized messages at the appropriate time.
[0080] Messages sent by the message selection system to each of the users' client devices can be opened via an application executing on the respective client device. The application executing on the client device can display messages from the message selection system to the user of the client device and prompt the user to take an action or elicit a user response. The application can be configured to feed information back to the message selection system including data corresponding to when the message was received at the client device, when the application displayed the message, the activity the user performed on the application, the activity the user performed on the client device, etc. The message selection system can use the information collected or received from the users' client devices to evaluate the effectiveness of messages in persuading users to take an action or the content and timing of user responses, improve personalization of messages sent to individual users, and in some embodiments, monitor users and their behavior in response to messages.
[0081] In certain embodiments, the message selection system can maintain user profiles for multiple users. For each user profile, the message selection system can execute an invoker that can invoke a message object at a predefined time interval. Each message object can generate message candidates using message templates that are candidates for transmission to a client device associated with the user profile. A message object evaluator of the message selection system can incorporate the context data into a model of the message object and output a confidence value associated with the message candidate generated by the message object. The confidence value can indicate the likelihood that the message candidate sent to the user will have an intended effect. The message object evaluator can output a confidence value for the message object that satisfies certain constraints. The message object evaluator can evaluate the confidence value associated with each of the message candidates based on certain conditions, update the confidence value based on a cooldown factor, and decide to transmit the message candidate to the client device if the updated confidence value exceeds a predefined threshold.
[0082] The message selection system can receive response data from a reporting agent of an application executing on a client device associated with the user profile. A response evaluator of the message selection system can incorporate the response data into a model of the message object to improve the performance of the model based on the response data corresponding to a previous message sent to the client device. The response data can be compared to a target value (or desired endpoint) of the message object to improve the model of the message object. In this manner, the message selection system can learn from the user's message preferences and responses to improve the selection process of future message candidates as well as improve the quality of the confidence values calculated for future message candidates. In doing so, the message selection system can select and send fewer messages to the user client device using message templates that are candidates for sending to the client device associated with the user profile while increasing the likelihood that the user will engage with or respond to the message and achieve the desired endpoint of the message.
[0083] In accordance with at least one aspect of the present disclosure, a method can include identifying, by one or more processors, a plurality of users and a plurality of message objects. The method can include retrieving, by the one or more processors, context data for each of the plurality of users for the one or more message objects. The method can include generating, by the one or more processors, message candidates based on the one or more message objects and the context data. The method can include updating, by the one or more processors, a confidence value for each of the message candidates based on a cooldown factor to generate an updated confidence value. The method can include selecting, by the one or more processors, the message candidate based on the updated confidence value. The method can include transmitting, by the one or more processors, the selected message candidate based on the updated confidence value exceeding a predetermined threshold.
[0084] 2 illustrates a block diagram of an environment 200 for competitive message selection. The environment 200 includes a message selection system 202 and at least one remote computing device 270. The message selection system 202 can select one or more messages to send to the remote computing device 270 over the network 104 and receive data from the remote computing device 270, including data corresponding to actions taken in response to the one or more messages sent to the remote computing device 270.
[0085] The environment 200 may include multiple remote computing devices 270. The multiple remote computing devices 270 may belong to the same user or may belong to different users or entities of the message selection system 202. Each remote computing device may include an application 272 configured to communicate with the message selection system 202. The application 272 may include a reporting agent 274 configured to provide data corresponding to actions performed on the computing device in response to receiving a message from the message selection system 202. In some embodiments, the application 272 may be configured to provide data to the message selection system 202 as well as display messages generated and transmitted by the message selection system 202. Additional details regarding the application 272 and the reporting agent are provided below.
[0086] Referring now to the message selection system 202, the message selection system 202 includes a memory 204. The memory 204 includes a plurality of message objects 206. Each of the plurality of message objects 206 can include a message template 208, a model 210, constraints 212, and a goal value 214. The memory 204 also includes context data 220, which can include at least one user profile 222 and historical data 224. The memory can also store message candidates 240 and respective confidence values 242 for each message candidate 240. As described herein, the message selection system 202 can generate the message candidates 240 using the message objects and further calculate the respective confidence values 242 for the message candidates 240.
[0087] The message selection system 202 also includes an invoker 250, a message object evaluator 252, and a threshold manager 254 to evaluate each message candidate 240. The invoker 250 can be an application, applet, script, service, daemon, routine, or other executable logic for selecting, identifying, and invoking the multiple message objects 206. The invoker 250 can initiate an invocation process that is initiated by a cron job scheduled at regular intervals. Cron is a time-based job scheduler in Unix-like computer operating systems and can be used to schedule jobs (e.g., commands, or shell scripts). The invocation process can be performed based on time (e.g., the invocation process can be performed at 8 a.m. on Mondays, the invocation process can be performed every 30 minutes, the invocation process can be performed every other Thursday, etc.). The invocation process can be performed based on events (e.g., the invocation process can be performed when the message selection system 202 receives data from a remote computing device 270, as described below). The invoker 250 can initiate a call process for a particular group of users. For example, the invoker 250 can initiate a call process for a subset of users using a given function of the application 272. The invoker 250 can initiate different call processes for different users. For example, the invoker 250 can initiate a first call process at a first time for a first user and a second call process at a second time for a second user. The invoker 250 can be broken down into multiple asynchronous requests, such as getting active users, getting user data, writing call records to a database, etc.
[0088] An example of the invocation process can be shown in the following pseudocode:
number
[0089] The call process as illustrated by the pseudo code above can be performed at regular time intervals (e.g., every 30 minutes). The call process can include determining (e.g., getData(u)) each user's context data 220 (described in detail herein) and determining the current time (e.g., Date.now()). The call process can include writing a call record to a database (e.g., db.write(u.id,ts,hash(d)), which can include a hash of the context data 220. The call record can include a user identification, a timestamp, and data for each user (e.g., application data and user interest data). The application data can include user actions in response to messages received at the application. The user interest data can include user acceptability throughout the day, such as how often the user responds to messages received at the application. The call record can be stored in Amazon Simple Storage Service (AWS S3) as illustrated in the pseudo code above.
[0090] Identifying active users (e.g., users of an application, users responding to messages) can include querying the application 272 to obtain active users who have enabled messaging. Obtaining user data can include obtaining data for each user, which can include application data and user attention data. This data can be used to filter message objects 206 that do not satisfy the hard constraints. Writing the call records to a database can be performed to group future evaluations and to track the decision history of the message objects 206 for future model training. For example, the call record timestamp, the user's application data, and the user attention data can be used by the response evaluator 256 as inputs to the message object model 210. A hash of the context data 220 can be stored in the remote object for storage, which can save space in the database of the message objects 206. By storing some data remotely, the message selection system 202 can avoid performance degradation as a result of scarce disk space. The message object evaluator 252 can offload read request traffic from the message objects 206 to a remote database (e.g., AWS S3) during evaluation. The message selection system 202 can store user data through the use of a hash table. The hash table can store unique changes of the context data 220, which is space efficient, e.g., avoiding storage of duplicate data. Storing the data in a remote database can reduce the frequency at which the invoker 250 must copy data from a live database. Instead, the invoker 250 can read directly from the remote database during offline training of the model 210.The invoke architecture can include parallel user processing, which can include multiple processes for obtaining user data.
[0091] As described above, the environment 200 can include one or more remote computing devices 270. More specifically, the remote computing devices 270 can receive messages from the message selection system 202. The remote computing devices 270 can be Internet-enabled remote devices. For example, the remote computing devices 270 can be Internet of Things ("IoT") devices. IoT devices can include remote monitoring devices including remote motion sensors, temperature sensors, soil condition sensors, or video cameras. In some implementations, the remote computing devices 270 can be mobile computing devices (or components thereof), such as mobile phones, smartphones, laptops, or tablet computers. The remote computing devices 270 can have limited computational resources. The remote computing devices 270 can be woken, pinged, messaged, or otherwise contacted at predetermined intervals to measure, perform activities, or provide responses to requests for information. In some implementations, the remote computing devices 270 can receive requests at regular, predetermined intervals.
[0092] The remote computing device 270 may include an application 272 to enable the remote computing device 270 to communicate with the message selection system 202. For example, the application 272 may receive messages from the message selection system 202 and return responses to the messages to the response evaluator 256. The application 272 may receive messages when the remote computing device 270 receives data packets from the message selection system 202 over the network 104. When the message selection system 202 sends a message to the application 272, the application 272 may launch on the remote computing device 270. The application 272 may ping, wake, message, or otherwise send notifications to the remote computing device 270 to contact a user of the remote computing device 270. The application 272 may launch and display messages from the message selection system 202. The application 272 may include a reporting agent 274. The application 272 may include a graphical user interface for engaging a user of the remote computing device 270. In some embodiments, application 272 may include a smoking cessation application or any other digital therapeutic application capable of sending messages to a subject using the digital therapeutic application.
[0093] The reporting agent 274 of the application 272 may be an application, applet, script, service, daemon, routine, or other executable logic configured to monitor activity within the application 272. The activity may include one or more user-performed actions. For example, the user-performed actions may include breathing exercises, meditation sessions, yoga sessions, walking sessions, jogging sessions, running sessions, communication with colleagues, phone sessions, communication with medical advisors, or any other actions that may be detected, determined, or inferred by the reporting agent 274. In some implementations, the reporting agent 274 may monitor when an activity is started, when an activity is completed, when an activity is interrupted by another computational function occurring in another process of the computing device, or when an activity is paused or stopped by an interaction with the application 272's environment. For example, the application 272 may use acceleration data from the remote computing device 270 to detect the start of a walking session and the completion of a walking session. The application 272 can relay or otherwise provide data to the reporting agent 274 that a user of the remote computing device 270 has initiated and completed a walking session. The application 272 can correlate the data on the remote computing device 270 with responses generated by the user. For example, the application 272 can display information that requests a response from the user. The application 272 can display a message asking the user to make a selection on the application 272 to confirm that the user has completed the activity.The application 272 may display a message requesting the user to make a selection on the application 272 to confirm that the user did not complete the activity. The application 272 may display a message requesting user feedback related to the appropriateness of the message. For example, the user may make a selection on the application 272 to notify the reporting agent 274 that a message sent during an event scheduled on the user's calendar is inappropriate. The user may make a selection on the application 272 to notify the reporting agent 274 that a message sent after a meeting scheduled on the user's calendar is appropriate.
[0094] The remote computing device 270 can assign a time stamp to the activity. The reporting agent 274 can access data related to the activity and the time stamp related to the activity. The reporting agent 274 can detect that a user is accessing data on the remote computing device 270 to perform an activity. The reporting agent 274 can receive reports from the application 272 detailing the activities performed by the user. In some implementations, the reporting agent 274 can generate one or more messages to send reporting data to the message selection system 202.
[0095] Referring again to message objects 206, the message objects 206 may be a data structure that may be a framework for message-centric optimization. The message object data structure may include message templates 208, models 210, constraints 212, and objectives 214. An invoker 250 may invoke multiple message objects 206. The message objects 206 may generate message candidates 240 based on the message templates 208 that are evaluated by a message object evaluator 252.
[0096] The message object 206 can include one or more message templates 208. The message template 208 can include a text string. The message object 206 can use the context data 220 to populate the fields of the message template 208. The message object evaluator 252 can use the message template 208 to generate the message candidates 240. The message template 208 can be static text, a template, or instructions for generating a message string. The template can include instructions to retrieve data from a data library stored in the memory 204. The data from the data library can be substituted into the template before it is sent to the user. For example, the template can include placeholders or variables that are filled in or substituted into the template before the message template 208 is set on the remote computing device 270. For example, the message can include images, videos, processor executable instructions that are sent to and executed on the remote computing device 270, or metadata. The message template 208 of the first message object 206 can be the same or different from the message template 208 of the second message object 206.
[0097] The model 210 of the message object 206 can receive the context data 220 as input and output a confidence value 242 indicating the predicted effectiveness of the message to achieve the goal value 214. For example, the context data 220 can be an input to the model 210 of the respective model object 206. The context data can include user-specific data, date, time, day of the week, weather-related data, and the like. The model 210 can receive inputs and output various outputs. For example, the inputs can include historical data 224 or user data, and the outputs can include a success rate or confidence value 242 calculated based on the input data. For example, the model can correlate the historical data 224 specifying timestamps of the user's responses to the effectiveness of the message. The response evaluator 256 can update the model 210 based on the historical data. For example, if the response evaluator receives data from the reporting agent 274 indicating that a user of a remote computing device 270 running an application 272 does not respond to messages on weekends, the response evaluator 256 can update the model 210 to reduce the probability of sending messages on weekends. The model 210 can receive updates to learn from past data. The response evaluator 256 can update the model 210 used to generate messages for responses generated based on certain demographic time patterns based on user application data. For example, the response evaluator 256 can determine that users over a certain age are prone to dehydration on hot days. The response evaluator 256 can update the model 210 to send a message encouraging the user to stay hydrated.
[0098] The model 210 can be a machine learning model (e.g., a reinforcement learning model, a k-nearest neighbor model, a backpropagation model, a q-learning model, a genetic algorithm model, a neural network, a supervised learning model, or an unsupervised learning model). The model 210 can include a feature set that is a feature available to the machine learning model. The feature set can include user application data, temporal context, effectiveness of similar messages, and a user attention metric, which is described below. Retraining the model 210 with new data or time series can improve the model 210 associated with the message object 206. The response evaluator 256 can add new features to the model 210, which can include updating a data service to return more user context information. The response evaluator 256 can fill in the data points with new feature values at each historical timestamp, which can include recalculating the historical user context, allowing the response evaluator 256 to retrain the model on all historical data with the newly added features.
[0099] It should be appreciated that the message selection system 202 can use feedback received from the remote computing device 270 to which the message was sent to improve the message generation and selection capabilities of the message selection system 202. For example, the message selection system 202 can be configured to update the model 210 of the message object 206 used to generate the message candidates 240 based on historical data related to the message candidates 240 generated by the message object 206. The success or failure of the message object 206 in the past can improve the predictive capabilities of the model 210 in the future. The model 210 can use the sent messages and their corresponding results and feature sets in the learning phase of the model 210. Many different supervised machine learning models can be trained online or offline. The training of the model 210 can use a cross-validation approach to find optimal model parameters. Retraining of the message object model 210 can occur at each iteration (e.g., training cycle). Some models 210 (e.g., tree-based models, neural networks) can be trained on a rolling basis as new data is received, and existing models can incorporate new data without retraining using the entire data set. The message object 206 may include a model 210 that predicts a confidence value 242 of a candidate message 240 .
[0100] The message object 206 may include a set of constraints 212. The constraints 212 may include limitations or restrictions that dictate whether a message candidate 240 should be generated from the message object 206 or whether the generated message candidate 240 may be sent. The constraints 212 may include a predetermined time or time window for sending the message candidate 240. The constraints 212 may include a determined time and day of the week that defines a time window during which the message template 208 may be sent. The constraints 212 may include a user type (e.g., gender, occupation, company, or affiliation). The constraints 212 may include regular days of the week for sending the message. For example, the constraints 212 of the message object 206 may be a time-based window from 10:00 a.m. to 4:00 p.m. during which the message object evaluator 252 may send the message. When the message object evaluator 252 processes a message object 206 with the above constraints, the message object evaluator 252 can determine whether to generate a message candidate 240 (or whether a message candidate 240 generated from the message object 206 can be sent) based on the current time. The message object 206 can send messages only on certain days of the week (e.g., only on Mondays or Fridays). The constraints can include days or times that the message object evaluator 252 prohibits the message object 206 from sending, such as weekends. The message object 206 can have a constraint 212 that a message can be sent on Monday unless Monday is a holiday, in which case the message object 206 is restricted from sending messages. The constraints 212 can include information about holidays, leap years, time of day, time of week, time of month, time of year, or other calendar-related subjects. The constraints 212 can specify specific times or time windows during which the message object evaluator 252 can send messages. The constraints 212 may include a minimum confidence value 242 .
[0101] The message object 206 may include a goal value 214. The goal value 214 may correspond to an action (e.g., to have a user perform a particular action in the application 272). The goal value 214 may correspond to a reply or response. The goal value 214 may include a numerical value corresponding to the effectiveness of the message received by the remote computing device 270. The goal value 214 may correspond to an action in the application 272 and a desired percentage of completion of the action. For example, a first action in the application 272 may be a breathing exercise and a second action in the application 272 may be a jogging exercise. The goal value 214 corresponding to the first action may be 80% and the goal value 214 corresponding to the second action may be 15%. For example, a message that invites a user to perform a task in the application 272 may be assigned a numerical value corresponding to the number of times the user performed the task as a percentage of the number of messages sent to the user to perform the task within a certain period of time. Additionally, messages that include a subject matter inviting the user to perform a particular task in an application may be assigned a numerical value corresponding to the time that the user performed the task after the message was sent to the user. In another example, a message that invites the user to reply may be measured by determining whether the user replied to the message or whether the user did not reply to the message. The target value 214 for a rhetorical message may not be easily defined because rhetorical messages do not elicit a measurable user response.
[0102] The failure of the message or its ability to elicit the intended action can be used as input to the message object model 210. The first message object 206 can have a goal value 214 of getting the user to stay hydrated. The first message object 206 can have a message template 208 of "Hello [name], have a cup of tea". The second message object 206 can have a goal value 214 of getting the user to stay hydrated. The second message object 206 can have a message template 208 of "Hello [name], have a cup of iced water". The message object evaluator 252 can use context data 220 related to the outside temperature and the user's tendency to forget to stay hydrated to determine that the first message object 206 and the second message object 206 should generate message candidates 240. If the outside temperature is below 20° F., the first message object 206 may generate a first message candidate 240 with a high confidence value 242, as compared to the second message object 206, which may generate a second message candidate 240 with a low confidence value 242. If the message object evaluator 252 recently sent a message related to staying hydrated, the message object evaluator 252 may lower the confidence score 242 of the first message candidate 240 and the confidence score 242 of the second message candidate 240. If the first message candidate 240 has an updated confidence score 242 that passes the threshold, the message object evaluator 252 may populate the message template 208 with data from the user profile 222 and send the message candidate 240 to a user of a remote computing device 270 running an application 272.
[0103] Referring again to the message object evaluator 252, the message object evaluator 252 can generate message candidates 240. The message object evaluator 252 can generate or assign a confidence value 242 for each message candidate 240 generated from the message object 206. The message object evaluator 252 can use the confidence value 242 associated with the message candidate 240 to determine whether to refrain from sending a message to a user of the remote computing device 270 or to send one or more messages to a user of the remote computing device 270. The message candidate 240 can have a confidence value 242 higher than a threshold. The message candidate 240 can be a message generated from execution of the message object 206 by the message object evaluator 252. The message candidate 240 can include a message having a confidence value 242 lower than a threshold. The message object evaluator 252 can generate the message candidates 240 by passing the context data 220 (and other data) the message object 206. The message objects 206 may output or otherwise generate message candidates 240 if the data provided to the respective message objects 206 satisfies the constraints 212 of the message objects 206. The message object evaluator 252 may input the context data 220 into the message object model 210 to generate confidence values 242 for the message candidates 240. The message object evaluator 252 may update the confidence values 242 based on a cool-down factor to generate updated confidence values 242. The message candidates 240 may be selected for transmission to the remote computing device 270 based on the updated confidence values 242. The message candidates 240 may be generated by the message object evaluator 252.
[0104] The message candidate 240 may include a confidence value 242. The confidence value 242 may be normalized and may include a confidence value 242 between 0 and 1 (inclusive) indicating the likelihood that the message will have the intended effect on the user of the remote computing device 270. The confidence value 242 may include other ranges, such as between 0 and 100 (inclusive). The relevance of a message may include a combination of message content, user data, user attention, and temporal context. Estimating the effectiveness of a particular message for a particular user at a particular time may include a supervised machine learning approach. The current temporal context (e.g., time of day, day of week, or week of year), user application data, and user attention may be used as a function of the supervised machine learning model. The type of decision model (e.g., polynomial regression, decision tree, or pairwise comparison analysis) may vary for each message object 206. The model 210 for each message object 206 may output a confidence value 242 indicating the confidence of the target value 214. The message object evaluator 252 can update the trust value 242 when the invoker 250 begins the invocation process.
[0105] It should be appreciated that the message selection system 202 can use the context data 220 as input to the model 210 for generating the confidence value. The context data can include a user profile 222 and historical data 224. The context data 220 can include the effectiveness of similar messages. For example, the message object 206 can use the context data 220 as input for generating the confidence value 242. The message object 206 can use the context data 220 to evaluate the constraints 212. The similar messages can be predictive indicators of the success or failure of the current message under scrutiny. The similar messages can be grouped by purpose or subject. The purpose of the message can include a call to action (to get the user to engage with the functionality of the application 272) or a call to reply (to get the user to respond to the message). The subject of the message can be, at a high level, functional (e.g., mission), informational, health, or financial. The context data 220 can be generated manually or by an unsupervised machine learning approach.
[0106] The contextual data 220 can include a user profile 222. The user profile 222 can include data related to a particular user. The user profile 222 can include a digital representation of the user's identity, such as the user's name, location, historical data, habits, application usage, application interactions, and response rates. The user profile 222 can include data derived from use of the application 272. The data can include user event data specific to the application 272, such as the number of times the user smoked a cigarette and the timestamps when the user smoked a cigarette. This data can be important for applications including smoking cessation. The data can include finer granularity of user engagement data, such as when the application 272 was opened. In some implementations, the message selection system 202 can generate derived functions from the raw data, depending on the type of model 210 implemented. For example, the derived function can include a list of timestamps of the user's cigarette smoking events, rather than raw data of the number of smoking events per day.
[0107] The context data 220 may include historical data 224. The historical data 224 may include a current time. The historical data 224 may include a temporal context. The temporal context may include available information about the current time, such as the time of day adjusted from the time zone, the day of the week, and the season. The historical data 224 may be generated by a reporting agent 274. The reporting agent 274 may assign a timestamp to a user action or response. The reporting agent 274 may transmit the historical data 224 to the response evaluator 256 over the network 104. The reporting agent 274 may assign a timestamp when the message was received by the remote computing device 270. The reporting agent 274 may assign a timestamp when the message was received by the user of the remote computing device 270.
[0108] The context data 220 can include a user tension indicator. User tension can include a user's receptiveness throughout the day, which can be a factor in whether a message has the intended effect. User attention can be highly specific. The message object evaluator 252 can identify various features of user tension to input into the model 210 for training the message objects 206. The message object evaluator 252 can use an initial guess, historical user time of day preferences, and application engagement data as input into the model 210 for training the message objects 206. The initial guess can include an initial hypothesis regarding user tension. When a user first signs up for the application 272, there may not be any historical data to work with, so the message object evaluator 252 can make an initial hypothesis randomly or based on human guessing. The human guess can be made based on data from other users. For example, a user may interact with the application 272 primarily in the mornings or evenings on weekdays. The best human guess may be based on demographic data. For example, if there are clear patterns, the response evaluator 256 can create clusters of users based on demographic data. The response evaluator 256 can use a time of day model as an initial hypothesis of user attention. The response evaluator 256 can plot the user's attention over a span of a day or a span of a week. Over time, the response evaluator 256 can historically display the user's time of day preferences as more messages 208 are sent to the user. As more messages are sent and results are received, a more accurate picture of the times when the user is likely to be receptive to messages can be built. The model 210 can weight the historical inputs based on the amount of historical input present in the database and the effect of success rate, which can increase the predictiveness of the model 210.The response evaluator 256 can use a learning rate to control how quickly the message selection system 202 learns the user's time of day preferences. The response evaluator 256 can select a learning rate to avoid overfitting the engagement patterns and to learn the user's preferences as quickly as possible. Using a high learning rate can overfit the engagement patterns and capture random noise. Using a low learning rate can take longer than necessary to learn the user's time of day preferences. The application engagement data can include general time of day engagement data generated by the application 272. The response evaluator 256 can use the application 272 engagement data to predict the effectiveness of messages, thereby making the application engagement data a proxy for user tension. The context data 220 can be generated by the reporting agent 274. The reporting agent 274 can receive feedback from the user.
[0109] Referring back to the message object evaluator 252, more particularly, the message object evaluator 252 may be an application, applet, script, service, daemon, routine, or other executable logic for determining when to generate message candidates and for sending one or more of the message candidates 240. The message object evaluator 252 may pass the context data 220 to the message object 206 to generate the message candidates 240. The message object evaluator 252 may generate a confidence value 242 for the message candidates 240. For example, the message object evaluator 252 may take a set of confidence values 242 and compare the confidence values 242 to one or more thresholds, further described below, to determine whether the message candidates 240 should be sent. The message object evaluator 252 may determine which, if any, of the message candidates 240 should be sent to a user of the remote computing device 270 based on the confidence values of the message candidates 240.
[0110] The message object evaluator 252 can select a message candidate 240 from a pool of message candidates 240 generated by a particular invocation of a set of message objects 206 based on the confidence values 242 of the message candidates 240 that meet a predefined threshold. The message candidates 240 can be weighted by the confidence values 242. The message object evaluator 252 can adjust the confidence value 242 of the selected message candidate 240 by multiplying the confidence value 242 by a cooldown factor. The message object evaluator 252 can adjust the confidence value 242 of the selected message candidate 240 by adjusting the confidence value 242 by the cooldown factor. In some embodiments, the message object evaluator 252 can adjust the confidence value 242 of the selected message candidate 240 by scaling the confidence value 242 by the cooldown factor. The message object evaluator 252 can determine whether to send the selected message candidate 240 based on the confidence value 242 exceeding a threshold. For example, a confidence value 242 of 0.8 exceeds a threshold of 0.7. The message object evaluator 252 may send a message if the message object 206 satisfies the constraints 212. The constraints 212 may be as follows:
number
[0111] Each message object 206 (or message candidates 240 generated thereby) may compete directly with each other for a particular user at regular time intervals. Competition between message candidates 240 may include, for example, ranking the message candidates 240 based on confidence values 242 indicating the likelihood that the message will have an intended effect on the user of the remote computing device 270. The message object evaluator 252 may rank the message candidates 240 based on the confidence values 242. The message object evaluator 252 may rank the message candidates 240 from highest to lowest confidence values 242. The message object evaluator 252 may select the message candidates 240 for transmission based on the rank of the message candidates 240. In some implementations, the message objects 206 may include confidence values 242, and the message objects 206 may be sorted or ranked, for example, by the confidence values 242. The message object evaluator 252 may select one or more of the message objects 206 to generate the message candidates 240. In some implementations, the message object evaluator 252 ranks and sorts the message objects 206. The message object evaluator 252 can transmit the message candidates 240 when the message object evaluator 252 completes its evaluation of the message objects 206. In some implementations, the generated message candidates 240 can be ranked and sorted, and selected portions of the message candidates 240 can be transmitted to a client device or a remote computing device 270.
[0112] The conflict among the message objects 206 may include sorting the message objects 206, such as sorting the message objects 206 based on the confidence values 242 or selecting a subset of the message objects 206 based on the categories of the message objects 206. For example, the categories of the message objects 206 may include fitness workouts, breathing exercises, stress relief activities, hydration reminders, smoking cessation tips, etc. The message object evaluator 252 may select a subset of the message objects 206 related to fitness workouts to select a message candidate 240 that encourages a user of the remote computing device 270 to perform a workout. The message object evaluator 252 may select the message candidate 240 with the highest confidence value 242 generated by the message objects 206. The message candidate 240 with the highest confidence value 242 may be selected by the message object evaluator 252 for transmission to the remote computing device 270 if the associated confidence value 242 exceeds or meets a predefined threshold.
[0113] The message object evaluator 252 can perform an evaluation process of the message candidates 240. The evaluation process can include a process by which the message object evaluator 252 can make a decision to send or not send the message candidate 240. The message object evaluator 252 can generate the message candidates 240 by passing the context data 220 to the message object 206 and evaluating the constraints 212. For example, the message object 252 can take the context data 220 related to the user's smoking habits and the history data 224 related to the time of day. The message object 206 can generate the message candidate 240 to generate the message candidate 240 "Hi Jim! Instead of smoking a cigarette, try taking a walk." The constraints 212 of the message object 206 can be that the message is not appropriate during the user's typical sleeping hours. The user profile 222 can indicate that the user typically wakes up between 7:00 a.m. and 11:00 p.m., and the history data can indicate that the current time is 1:00 p.m. If the constraints 212 are satisfied, the message object evaluator 252 may generate a message candidate 240 and record a confidence value 242 in a database of message candidates 240. At regular intervals, the message object evaluator 252 may identify message objects 206 that have not yet generated a message candidate 240 and generate a message candidate 240 for the message object 206. The message object evaluator 240 may determine a confidence value 242 for the message candidate 240. The message object evaluator 252 may identify the message candidates 240, weight or sort the message candidates 240 by the confidence value 242, scale the message candidates 240 by a cool-down factor, and then select a message candidate 240. The message object evaluator 252 may use the confidence value 242 and the threshold policy 234 to determine whether or not to send the message candidate 240.For example, if the confidence value 242 is above a threshold according to the threshold policy 234, the message object evaluator 252 may decide to send the message candidate 240 to the remote computing device 270. If the confidence value 242 is equal to or less than a threshold according to the threshold policy 234, the message object evaluator 252 may decide not to send the message candidate 240 to the remote computing device 270. If the message object evaluator 252 makes a decision to send the message candidate 240, in some embodiments, the message object evaluator 252 communicates the message candidate 240 to a messenger service of the message selection system 202 or to the message selection system 202, which facilitates sending the message candidate 240 to the user's device or to the remote computing device 270. The reporting agent 274 may record the actions of the remote computing device 270 receiving the message candidate 240. If the message object evaluator 252 decides not to send the message, the message object evaluator 252 does not send the message.
[0114] Referring again to the response evaluator 256, the response evaluator 256 can analyze responses from users of remote computing devices 270. The response evaluator 256 can record the results of messages sent to users, such as the success or failure of the message object 206 to achieve the message object goal value 214. The response evaluator 256 can record the results of messages to obtain a goal value 214 for the learning algorithm to optimize. Each message object 206 can define the success of the message object 206 to achieve the goal value 214. The computation of the results can be implemented as a batch process or a real-time streaming process.
[0115] The response evaluator 256 can add new features to the model 210, which can include updating the data service to return more user context information. The response evaluator 256 can fill in the data points with new feature values at each historical timestamp, which can include recalculating past user context, and the response evaluator 256 can retrain the model on all historical data with the newly added features. The response evaluator 256 can create clusters of users based on demographic data. The response evaluator 256 can use a time of day model as an initial hypothesis of user attention. The response evaluator 256 can graph the attention of the user over a span of a day or a span of a week. The response evaluator 256 can historically understand the time of day preferences of the user as more messages 208 are sent to the user over time. The response evaluator 256 can use a learning rate to control the rate at which the message selection system 202 learns the time of day preferences of the user. The response evaluator 256 can select a learning rate to avoid overfitting engagement patterns and to learn user preferences as quickly as possible. The response evaluator 256 can use the application engagement data to predict the effectiveness of messages, thereby making the application engagement data a proxy for user interest.
[0116] The response evaluator 256 can use the actions recorded by the reporting agent 274 to determine a cool-down factor for future evaluations. The cool-down factor can be specific to an individual user of the application 270. The cool-down factor can reflect a variety of factors, including, but not limited to, the time since the last message was received by the remote computing device 270, the total number of messages the user of the remote computing device 270 chooses to receive within a given time period, and the length of time the user of the remote computing device 270 is awake or active. The cool-down factor can decrease the confidence value 242 of a message candidate 240 based on how recently the message was last sent to the user of the remote computing device 270. For example, at time t=t1, the message object evaluator 252 can send a message to the user of the remote computing device 270 encouraging them to engage in a fitness workout. At time t=t1+1, the message object evaluator 252 multiplies the trust value 242 of the message 2 candidate 40 generated at time t=t1+1 by a cooldown factor of 1 / 2 so that the updated trust value 242 is 1 / 2 of the trust value 242 before the update. At time t=t1+4, the message object evaluator 252 multiplies the trust value 242 of the message candidate 240 generated at time t=t1+4 by a cooldown factor of 5 / 8 so that the updated trust value 242 is 5 / 8 of the trust value 242 before the update, and the message object evaluator 252 is more likely to send the message at time t=t1+4. The response evaluator 256 can update the cooldown factor based on the user action taken in response to the message.For example, the response evaluator 256 can determine that a user is likely to experience message fatigue based on a current cool-down factor, and therefore the response evaluator 256 can increase the cool-down interval which can decrease the cool-down factor. The cool-down factor can be determined by one or more policies 230 described herein.
[0117] The message selection system 202 can store policies 230 for managing whether and when to send a message based on a calculation of the trust value 242. The policies 230 can include a cooldown policy 232 and a threshold policy 234. The cooldown policy 232 can include a cooldown factor. The cooldown factor can account for message fatigue by lowering the trust value 242 of each candidate message 240 relative to the recent state in which the last message was received by the remote computing device 270. For example, a linear cooldown rate with a cooldown interval equal to the user's wake-up time period w, which can be reported or monitored by the user via the application 272, divided by the maximum number r of messages the user chooses to receive in a day. The user can choose to receive a certain number of messages, or the message selection system 202 can determine the number of messages to send to the user without inducing message fatigue. If the user chooses r=4 and w=14 hours, the cooldown interval is 14 / 4=3.5 hours. If the cooldown interval is 1 hour, the cooldown rate is linear, and the last message was received 10 minutes ago, the cooldown factor may be 1 / 6, lowering the overall confidence value 242. The cooldown rate and cooldown interval may be adjustable parameters in the model 210. For example, the model 210 may be a linear model, an exponential model, or a decay model. The type of model may be important for the message object evaluator 252 to build the model 210 to predict the effectiveness of a message in eliciting a user response.
[0118] It should be appreciated that the goal value 214 may be used by the response evaluator 256 to train the model 210. The goal value 214 may represent whether the message had the intended effect, which may include a call to action (having the user use a feature within the application 272) or a call to response (having the user reply to the message). Application data may be used to monitor the effect and record the outcome of the intended effect. A binary value may be used as the goal value 214. One goal of the message selection system 202 is to promote the efficacy of the message object 206 in achieving the goal value 214. The call to action and the call to response may directly compete with each other. The intended effect may include success in equally valued categories. The intended effect may include success in differently weighted categories.
[0119] The message selection system 202 can include one or more threshold managers 254. The threshold manager 254 can be an application, applet, script, service, daemon, routine, or other executable logic for managing threshold policies 234. The threshold manager 254 can determine which threshold policy 234 to apply to a given user, a given context (e.g., time of day), or a message object 206. The threshold manager 254 can select from a variety of threshold policies 234, such as fixed thresholds, dynamic thresholds, weighted random thresholds, and fuzzy fixed thresholds. The threshold manager 254 can select or determine a cooldown rate, a cooldown interval, a last message interval, a cooldown factor, and a distribution control factor.
[0120] The policy 230 can include a threshold policy 234. The threshold manager 254 can control the quality of the messages sent, for example, by selecting candidate messages 240 with values greater than a threshold. The quality of the message can be a quantitative indication of the usefulness to the user or a quantitative indication of the likelihood that the user will act on the message. The threshold manager 254 can control the approximate number of messages sent to the remote computing device 270 received in a day. For example, given a desired number of messages per day, the threshold manager 254 can perform a calculation based on a history of predicted confidence values 242. The threshold manager 254 can set the threshold such that the number of predictions that exceed the threshold equals the number of messages desired to be delivered in a day. The threshold manager 254 can set the threshold by relaying the threshold to the message object evaluator 252. The message object evaluator 252 can retrieve the threshold, or thresholds, from the stored memory 204. Because the threshold is based on probability, the number of messages delivered in a day can be approximate. The exact number of messages delivered may not necessarily be the desired number of delivered messages. For example, the message selection system 202 may send high value messages even if sending high value messages would result in a greater number of messages sent than the desired number of delivered messages. Also, the message selection system 202 may not send low value messages even if sending low value messages would result in a greater number of messages sent than the desired number of delivered messages. The message object evaluator 252 may determine whether sending or refraining from sending a message would lead to message fatigue or positive user engagement with the application 272 based on the threshold policy 234. The distribution of confidence values 242 may be obtained from mathematical simulations or mocks performed during the invocation process by the invoker 250.The number of messages per day for a particular user is a parameter input to message object evaluator 252. The number of messages per day sent to a particular user can be a default number, such as 3, or it can be explicitly provided as a preference by the user, or it can be provided by another model 210 that has learned the optimal number of messages per day given the user's context.
[0121] The threshold policy 234 may include a fixed threshold. The fixed threshold may find a threshold on a given day that is any value above the threshold that indicates the optimal time to send a message. The threshold is calculated for a given day and does not change. The effect of a non-changing threshold is that the candidate messages 240 sent are effectively optimized. The message object evaluator 252 may not send messages frequently when the threshold policy 234 is a fixed threshold. FIG. 4 shows a graph 400 of confidence values versus time with a sharp peak. The graph 400 has a sharp peak 402. The sharp peak 402 may indicate that the message candidates 240 with high confidence values 242 for a set of message candidates are distributed toward a certain time. The message object evaluator 252 may not send a user-specified number of messages or a number of messages determined by the message selection system 202 if the messages with high confidence values 242 are clustered at the time 404. The message object evaluator 252 may adjust the confidence value 242 based on the cool-down factor, which may make it less likely that a message is sent at time 406, even though the message at time 406 has a high confidence value compared to the message at time 408. If the threshold is set at a confidence value 242 of 0.8 depicted by line 410, the message object evaluator 252 will not send a message at time 406 that has a confidence value 242 of 0.7. A certain threshold may prevent the message object evaluator 252 from sending a message at time 406, even though the message at time 406 has a high confidence value 242 compared to the message at time 408. A single peak 402 may indicate that a user has enough attention span for a few messages (e.g., 2 or 3) per day, and any more messages would cause message fatigue. However, the message object evaluator 252 may miss an opportunity to send a message at time 406 that has a high confidence value below the threshold line 410.
[0122] The threshold policy 234 may include dynamic thresholds. The threshold manager 254 may use dynamic thresholds if it is desired to send a minimum number of messages per day to a remote computing device. The threshold manager 254 may recalculate the thresholds based on the current time of day. Dynamic thresholds may include finding the next future peak or set of peaks, solving the problem of missing a firing window when confidence values are biased toward the beginning of the day. With dynamic thresholds, it is guaranteed that there is at least one time in the future that the message object evaluator 252 can send a message, which means that the user can receive the desired number of messages per day. Dynamic thresholds may cause the threshold manager 254 to wait until the end of the day to send a message and time out by sending the message based on a cooldown rate. Although dynamic thresholds may achieve control over the number of messages the message selection system 202 sends per day, the threshold manager 254 may choose a suboptimal time to send and may not make a decision regarding the number of messages to send per day. For example, the threshold manager 254 can recalculate the threshold line 410 for different times of the day. The threshold line 410 can be 0.8 between 0800 and 1000 hours, 0.9 between 1001 and 1800 hours, and 0.7 between 1801 and 2100 hours. The threshold can be highest during the user's work hours of the day and lower in the mornings and evenings. A dynamic threshold policy is useful in situations where a user receives fewer messages during work hours and more messages at the beginning and end of the day.
[0123] The threshold policy 234 may include a weighted random threshold. The weighted random threshold may include a probability of sending a message for a message candidate trust value 242. The message object evaluator 252 may scale the weight of the probability for the trust value 242, and the parameters to the algorithm are derived from data (by empirical distribution) such that the weighted random threshold sends approximately the desired number of messages per day. The probability-based threshold means that a message is allowed to be sent with a high probability at the trust peaks 402 and a low probability at the trust valleys 412. The advantage of the weighted random threshold is that the message object evaluator 252 may still send messages during non-peak times of the day, introducing more randomness into the decision by the message object evaluator 252. With more randomness means it becomes harder to control the number of messages sent per day. Changing the rate at which the invoker 250 activates the message objects 206 and the message object evaluator 252 evaluates the message objects 206 changes the number of messages sent to the user in a day if the probability weights are held constant. The message selection system 202 can recalculate the probability weight parameters and apply the recalculated probability weight parameters each time the message selection system 202 changes the invocation time interval to achieve the same number of messages sent in a day.
[0124] The threshold policy 234 can include a fuzzy constant threshold. Figure 5 shows a graph of confidence value versus time with upper and lower thresholds. The fuzzy constant threshold can include a value below an upper threshold 502 selected with a certain probability for the value in question. The lower threshold 504 can be selected to set the probability of sending a message to zero if the confidence value 242 is too low. For example, a threshold of 0.5 can be set to send a message with a probability of 1, a threshold of 0.15 can be set to send a message with a probability less than 1 and greater than 0, and a threshold of 0.1 can be set to send a message with a probability of 0. The fuzzy constant threshold can be similar to a constant threshold without a hard threshold, where above the confidence value 242 the message object evaluator 252 sends a message and below it the message object evaluator 252 refrains from sending a message.
[0125] The distribution control factor can be used to control how quickly the probability of sending a message drops relative to the distance below the threshold. The distribution control factor is a mechanism that controls the distribution of the number of messages per day. The larger the distribution control factor, the higher the probability that the message object evaluator 252 will send the desired number of messages per day. The fuzzy constant threshold can include the following algorithm: T=(CV DCF ) × CV, CV∈[0,1] (Equation 3) Here, CV is the confidence value, DCF is the distribution control coefficient, and T is the threshold value.
[0126] The fuzzy constant threshold can have the effect that the message object evaluator 252 sends message candidates 240 with low confidence values 242 below the upper threshold 502, but refrains from sending messages with confidence values 242 below the lower threshold 504. There is a chance of sending messages during non-peak times of the day, introducing some randomness into the message sending decision. The message object evaluator 252 may send message candidates 240 with confidence values 242 that do not exceed the threshold to the remote computing device 270. Sending messages below the threshold may be advantageous to draw user attention in situations not predicted by the model 210. By changing the rate at which the message objects 206 are invoked and evaluated, the number of messages sent to a user per day can be changed if the distribution control factor is held constant. The message selection system 202 may recalculate and apply the distribution control factor parameters each time the invoker 250 changes the invocation time interval, so that the number of messages sent per day remains the same.
[0127] 6A-6B are diagrams illustrating distributions of the number of messages sent in a day. FIG. 6A depicts a distribution of messages sent on a particular day with a mean of 3 and low variance, corresponding to a large number of messages sent around 3:00 on a particular day. FIG. 6B depicts a distribution of messages sent on a particular day with a higher variance than the distribution of FIG. 6A. Threshold manager 254 can apply a distribution control factor to the distribution of FIG. 6B to generate the distribution of FIG. 6C, which has a distribution control factor of 2. By applying a distribution control factor to the distribution, threshold manager 254 can reduce the variance of the distribution of the number of messages sent.
[0128] Edge cases may arise where the values around the threshold are highly repetitive. For example, a hypothetical list of trust values 242 could include the following: [0.95, 0.95, 0.95, 0.96, 0.96, 0.96, 0.96, 0.96, 0.96, 0.97]. If a threshold were drawn between the 6th and 7th elements of the list of trust values 242 (calculated based on messages / day=3), there would be a high probability of sending more messages in a day than desired, which would occur because there are many trust values 242 of 0.96 both above and below the threshold. A simple calculation that simply looks for values above the threshold may result in sending more messages 208. The calculation of message objects 206 can be highly repetitive. For example, if 24 invocations are evenly distributed throughout the day, but the user only uses application 272 at the beginning and end of the day, evaluation of message objects 206 during periods of non-use would calculate trust values 242 based on the same data and end up producing the same trust values 242. Because the message selection system 202 has data indicating repeating confidence values 242, the system can limit the number of messages sent to a user even if the confidence value 242 of a candidate message 240 is greater than a threshold.
[0129] FIG. 3 illustrates a flow diagram of an example method 300 for competitive message selection. The method 300 may include identifying, by one or more processors, a plurality of users and a plurality of message objects (block 302). The method 300 may include searching, by one or more processors, respective context data of the plurality of users for the one or more message objects (block 304). The method 300 may include generating, by one or more processors, message candidates based on the one or more message objects and the context data (block 306). The method 300 may include updating, by one or more processors, a confidence value for each message candidate based on a cool-down factor to generate an updated confidence value (block 308). The method 300 may include selecting, by one or more processors, a message candidate based on the updated confidence value (block 310). The method 300 may include determining whether the confidence value exceeds a predetermined threshold (block 312). The method 300 may include transmitting, by one or more processors, the selected message candidate (block 314).
[0130] As described above, the method 300 may include identifying a plurality of users and a plurality of message objects (block 302). Identifying the plurality of users may include accessing a database to determine active accounts. Also referring to FIG. 2, among other things, the reporting agent 274 may send a report to a response evaluator 256 that may analyze responses from the remote computing device 270. Identifying the plurality of users may include accessing user profiles 222 of the plurality of users. Identifying the plurality of message objects may include identifying a plurality of message objects 206 stored in the message selection system 202. Identifying the plurality of message objects 206 may include accessing a plurality of message objects 206 stored in the message selection system 202. The invoker 250 may query the application 272 to obtain active users for which messaging is enabled.
[0131] The method 300 may include retrieving context data for each of the plurality of users for one or more message objects (block 304). Retrieving the context data may include retrieving the context data 220 including the user profile 222 or the history data 224. The message object evaluator 252 may retrieve the context data to determine whether and when to send a message to the user. Retrieving the context data may include retrieving the context data for each of the plurality of users identified in block 302. Retrieving the context data may include accessing the context data 220 in the user profile 222 for each of the plurality of users identified in block 302. For example, for an identified user, the message object evaluator 252 may collect the context data 220 for the identified user by accessing the user profile 222 for the identified user. The invoker 250 may obtain data for each user, including application data and an indication of user tension. This data may be used to filter message objects 206 that do not satisfy the constraints 212.
[0132] The method 300 may include generating message candidates based on one or more message objects and the context data (block 306). The message object evaluator 252 may generate the message candidates. The message candidates 240 may be based on one or more message objects 206 and the context data 220. For example, the first message candidate 240 may be generated by a first message object 206 for the identified user. The second message candidate 240 may be generated by a second message object 206 for the identified user. The third message candidate 240 may be generated by a third message object 206 for the identified user. The first message candidate 240 may be generated based on the user profile 222 of the user and the user's time of day preferences. The second message candidate 240 may be generated based on the user profile 222 of the user and the availability of similar messages. The third message candidate 240 may be generated based on the temporal context of the message and the user application data. The message object 206 can generate a message and send it to a remote computing device and store it as a message candidate 240. The message candidate 240 can be generated and associated with a confidence value 242. The confidence value 242 associated with the message candidate 240 can indicate how confident the message object 206 is in achieving the message object goal value 214. The message object 206 can generate a first or original confidence value 242 associated with the message. The message object 206 can generate the message candidate 240 based on the constraints 212. The message object 206 can fail to generate the message candidate 240 because the message object 206 did not satisfy the constraints 212.
[0133] The method 300 may include updating a confidence value for each message candidate based on a cool-down factor (block 308). The threshold manager 254 may include updating the confidence value 242 for each message candidate 240 based on the cool-down factor to generate an updated confidence value 242. Updating the confidence value 242 may include accessing a cool-down policy 232 of the message selection system 202 and applying the cool-down policy to generate an updated confidence value 242 for the message candidate 240. Generating the updated confidence value 242 based on the cool-down factor may apply to the first confidence value or the original confidence value. The updated confidence value 242 may be stored separately from the first confidence value 242 generated by the message object 206. For example, in scenarios where different cool-down policies may be applied at different times, the original confidence value or the first confidence value may be stored along with any updated confidence value. The cool-down factor may be a scaling factor that updates the confidence value 242 for each message candidate 240.
[0134] The method 300 may include selecting a message candidate based on the updated confidence values (block 310). The message object evaluator 252 may select a message candidate 240, and the selected message candidate 240 may be sent to a remote computing device. The selected message candidate 240 may have a highest confidence value 242 of a set of message candidates 240 that may be sent to a user. The message candidate 240 may not have a highest confidence value 242 of a set of message candidates 240 that may be sent to a user. For example, a first message candidate 240 may have a lower confidence value 242 than a second message candidate 240, but the first message candidate 240 is sent to the user due to the threshold policy 234. The message object evaluator 252 determines when to execute the message object 206 to generate the message candidate 240 and when to send the selected message candidate 240. A message object 206 may decide not to send the message of the message object 206 even if the message candidate 240 has a confidence value with the maximum value in the pool of message candidates 240 .
[0135] The method 300 may include determining whether the confidence value exceeds a predetermined threshold (block 312). If the confidence value 242 exceeds or meets the predetermined threshold, the message object evaluator 252 may transmit the selected message candidate. If the confidence value does not exceed or meet the predetermined threshold, the system may identify multiple message objects 206 or select a new message candidate 240.
[0136] The method 300 may include sending the selected message candidate to the remote computing device 270 (block 314). The message object evaluator 252 may compare the updated confidence value 242 to a threshold to determine whether to send the selected message candidate 240. The threshold may be based on various threshold policies 234 described in the previous section.
[0137] After the message is sent, the remote computing device 270 can receive the message. The remote computing device 270 can display the message in an application 272. The user can receive the message, which can trigger a response or reply from the user. The user can receive the message, which can elicit an action from the user.
[0138] C. Indirect updates to messaging systems.
[0139] A messaging system may deliver messages that improve health, wellbeing, fitness, and overall quality of life. However, a messaging system may risk delivering an oversupply of messages to a user that may lead to message fatigue. A messaging system may use responses to messages to update a model for selecting when to send messages to customize the sending time and content of future messages. Messages sent by a messaging system may be non-rhetorical and rhetorical. A non-rhetorical message may prompt a user of an application for a response or other input. A rhetorical message may not prompt a user of an application for a response or other input. For example, a rhetorical message may provide information or data to a user. Without user response data on the effectiveness of a rhetorical message, a messaging system may not be able to update a model for selecting when to send a rhetorical message.
[0140] This disclosure describes systems and methods for indirectly updating a model of message objects in a messaging system. Because the messaging system may not receive responses to rhetorical messages, the messaging system uses responses to non-rhetorical messages to update a model of message objects associated with the non-rhetorical messages. Because rhetorical messages do not implicitly generate a response from a user, the messaging system cannot determine the effectiveness or ineffectiveness of a rhetorical message based on the responses to the rhetorical message. Furthermore, without a response from a user, the messaging system may not be able to quantify the benefit that a rhetorical message provides to a user.
[0141] To address some of the challenges associated with the lack of response data from rhetorical messages, the present disclosure relates to indirect updates to a model of message objects in a message selection system. The message selection system can include a response evaluator that can incorporate response data associated with non-rhetorical messages to update a model of message objects associated with rhetorical messages. In this manner, the message selection system can use the response data of the non-rhetorical messages as a proxy for the response data (or lack thereof) of the rhetorical messages. In this manner, the message selection system can select and send a customized rhetorical message to a user's client device without receiving direct feedback regarding the rhetorical message. The message selection system can also send the rhetorical message to the user's client device at a customized time to prevent message fatigue.
[0142] 7 is a block diagram of an environment 200 of a system for sending a message to a user. The environment 200 includes a number of modules, components, or other entities executing within a message selection system 202. The components of the message selection system 202 may include any type and form of application, program, library, script, service, process, task, or executable instruction executing on a device, such as a mobile device, computer, or server. The message selection system 202 may include any of the other message selection system 202 components described herein, such as those described in connection with FIG. 2. In addition to the components described in connection with the message selection system 202 described in connection with FIG. 2, the message selection system 202 may include a similar message object selector 712 and a response classifier 714.
[0143] The message selection system 202 can include a number of message objects 206. The message objects 206 can be non-rhetorical message objects 206 or rhetorical message objects 206. The non-rhetorical message objects 206 can have message templates 208 from which non-rhetorical messages are generated. A non-rhetorical message is a message that elicits a response from a user of a remote computing device 270 when the message is received by the user of the remote computing device 270. For example, a non-rhetorical message can include processor-executable instructions that, when executed by the remote computing device 270, cause the remote computing device 270 to display, for example, a prompt or input to the user. Activation of the prompt or input can generate a response message that the remote computing device 270 sends to the message selection system 202. The non-rhetorical message objects 206 have a non-rhetorical message model 210. The model 210 can be updated based on responses from the remote computing device 270 generated in response to the non-rhetorical message. The non-rhetorical message object 206 has a goal value 214 that corresponds to a response or reply from a user of a remote computing device 270. The goal value 214 may correspond to an action.
[0144] The message objects 206 may also include rhetorical message objects. The rhetorical message objects 206 may have message templates 208 from which rhetorical messages are generated. Rhetorical messages may be messages that do not elicit a user response when the message is received by a user of the remote computing device 270. For example, a rhetorical message may encourage a user to drink more water without prompting the user to record their water intake or asking the user to confirm that they have increased their water intake. In another example, a rhetorical message may display information or data to a user when rendered by an application 272 at the remote computing device 270. The rhetorical message objects 206 have a rhetorical message model 210. The rhetorical message objects 206 have a goal value 214 that may correspond to an action. The message selection system 202 may determine the goal value 214 associated with the rhetorical message object 206 based on responses from the remote computing device 270 to the non-rhetorical message.
[0145] The message selection system 202 may include a response classifier 714. The similar response classifier 714 may be an application, applet, script, service, daemon, routine, or other executable logic for classifying responses from non-rhetorical messages. The response classifier 714 may search a database of responses in the message selection system 202 to generate a feature set for training or updating the model 210 of the message object 206. The feature set may include information related to a quality score of a response to a message sent from the message selection system 202. The feature set may include a time series indicating when the message was sent to the remote computing device 270, when (or if) a response to the message was received, and a quality score of the response. The quality score may indicate a level of user engagement with the received message. The response classifier 714 may determine a quality score of the response based on the time from sending the message to the remote computing device to receiving the response to the message by the message selection system 202. For example, the response classifier 714 may assign a high quality score to responses received within a predetermined duration from sending the message and a low quality score to responses received after the predetermined duration. In some implementations, the quality score may be based on a time decay function. In some implementations, the response classifier 714 may generate a quality score based on the content of the response. For example, if the message includes an input box for a user to enter text, the response classifier 714 may assign a quality score based on the amount of text entered into the input box.
[0146] The response classifier 714 can classify the response from the non-rhetorical message according to the response policy 716. The response classifier 714 can indicate whether the quality score is calculated based on the response time, the amount of interaction or engagement with the message (e.g., the amount of text entered into an input box), or a combination thereof. The response policy 716 can indicate that the quality score is time-based. For example, the response policy 716 can be based on the elapsed time between the receipt of the non-rhetorical message by the remote computing device 270 and the recording of the user's response by the reporting agent 274. For example, the remote computing device 270 can receive the message at 7:00 a.m. The non-rhetorical message can prompt the user of the remote computing device 270 to do five minutes of stretching before the user leaves for the office. The non-rhetorical message can include a button that prompts the user to confirm that the user engaged in a five-minute stretching session. The reporting agent 274 may receive a confirmation that the user engaged in a five minute stretching session at 7:10 a.m. shortly after the user received the message, indicating that the user engaged in a stretching session between 7 a.m. and 7:10 a.m. The response policy 716 may rank responses to the non-rhetorical message on a time basis. For example, the response policy 716 may rank a quick response to a non-rhetorical message higher than a slow response to a non-rhetorical message. The response policy 716 may include a time limit for the user to respond to the message.
[0147] The response policy 716 can indicate that the quality score is engagement-based. For example, a high quality score can be a response that has a high level of active user engagement. The message can prompt the user to answer several questions related to the content of the message, such as "How would you rate the activity?", "Would you like another reminder to drink water?", etc. Responses with active user engagement can include responses in which the user answers most of the questions prompted by the message.
[0148] The message selection system 202 may include a similar message object selector 712. The similar message object selector 712 may be an application, applet, script, service, daemon, routine, or other executable logic that selects a message object 206 based on the similarity of the message object to another message object 206. The similar message object selector 712 may identify a rhetorical message object 206 to update. The similar message object selector 712 may select a rhetorical message object 206 to update if the response evaluator 256 has not updated the model 210 within a predetermined time. For example, the similar message object selector 712 may update the model 210 daily, weekly, or monthly. The similar message object selector 712 may determine a classification of the rhetorical message object 206. The similar message object selector 712 may determine a classification of the rhetorical message object to find a similar non-rhetorical message object 206 that can be used to update the model 210 of the rhetorical message object 206.
[0149] The similar message object selector 712 can identify non-rhetorical message objects 206 that are similar to the rhetorical message object 206 to be updated. A feature set (e.g., response data) of the non-rhetorical message object 206 can update a rhetorical message object 206 of a similar classification. The similar message object selector 712 can identify non-rhetorical message objects 206 that are similar to the rhetorical message object 206 to be updated. A non-rhetorical message object 206 can be similar to a rhetorical message object 206 if the non-rhetorical message object 206 has the same classification, tag, or category as the rhetorical message object 206. For example, the similar message object selector 712 can identify a tag (e.g., breathing techniques) of the rhetorical message object 206 to be updated and search for one or more non-rhetorical message objects 206 that have the same tag. The similar message object selector 712 can identify message objects 206 that have the same tag. For example, a message object may have a "breathing techniques" tag or a "hydration" tag. The similar message object selector 712 may select a non-rhetorical message object 206 based on the similarity of the non-rhetorical message object to the rhetorical message object 206. The similar message object selector 712 may identify groups of message objects 206 that have the same tag or classification.
[0150] Referring back to the response evaluator 256, the response evaluator 256 receives information and data from the reporting agent 274 via the network 104 and generates updates to the model 210 of the message object 206. As described in the previous section of this disclosure, the model 210 of the message object 206 can receive the context data 220 as input and output a confidence value 242 indicating the predicted effectiveness of the message to achieve the target value 214. The model 210 can be a machine learning model (e.g., a reinforcement learning model, a k-nearest neighbor model, a backpropagation model, a q-learning model, a genetic algorithm model, a neural network, a supervised learning model, or an unsupervised learning model). The response evaluator 256 can train the model 210 based on a feature set generated from past responses to the message. The model 210 of each message object 206 can output a confidence value 242 indicating the confidence of the target value 214.
[0151] The response evaluator 256 can update the model 210 of the message object 206. The message object 206 to be updated can be selected by the similar message object selector 712 based on the classification by the response classifier 714. The response evaluator 256 can update the model 210 of the rhetorical message object 206 based on the response classified by the response classifier according to the response policy 716. For example, the similar message object selector 712 can identify non-rhetorical message objects 206 that have the same tag or classification as the rhetorical message object 206 to be updated. The similar message object selector 712 can identify response messages received by the message selection system 202 in response to messages from the identified non-rhetorical message object 206 and generate a set of functions that the response evaluator 256 can use to update the model 210 of the non-rhetorical message object 206 and the rhetorical message object 206.
[0152] The response evaluator 256 can update the model 210 of the message object 206 by training or retraining the model 210. The response evaluator can train the model 210 of the rhetorical message object 206 based on a feature set from the non-rhetorical message object 206. As described above, the response evaluator 210 can update the model 210 of the non-rhetorical message object 206 based on a response received to a message generated by the non-rhetorical message object 206. The response evaluator 256 can update the model 210 of the rhetorical message object 206 based on the model 210 of the message object 206 identified by the similar message object selector 712. For example, the response evaluator 256 can import or copy the model 210 from one of the identified non-rhetorical message objects 206 to the rhetorical message object 206 to be updated.
[0153] The response evaluator 256 can update the model 210 of the rhetorical message object 206 based on responses to the messages generated by the non-rhetorical message object 206. For example, the similar message object selector 712 can identify a plurality of non-rhetorical message objects 206 that have the same classification as the rhetorical message object 206 to be updated. The response evaluator 256 can identify a plurality of past messages and responses to the identified non-rhetorical message object 206. The response evaluator 256 can generate a feature set based on the plurality of past messages and responses to the identified non-rhetorical message object 206. The response evaluator 256 can then train a model of the rhetorical message object 206 based on the feature set generated from the plurality of past messages and responses to the identified non-rhetorical message object 206.
[0154] The message selection system 202 can determine the effectiveness of the rhetorical message. The response evaluator 256 can conduct an A / B test (e.g., split test or bucket test) in which a subset of users of the application 272 receives the rhetorical message and a subset of users of the application 272 does not receive the rhetorical message. In this manner, the message selection system 202 can measure the effectiveness of the rhetorical message. For example, a subset (e.g., half) of users (group A) can receive the rhetorical message and a subset (e.g., half) of users (group B) can not receive the rhetorical message. The users in group A can be different from the users in group B, and group A can include different people who receive the rhetorical message at the same time. The message selection system 202 can hold other variables constant (e.g., number of messages sent, time of day the messages are sent) to determine the effect of the rhetorical message on users' engagement with the application 272 and whether the rhetorical message leads to user behaviors that improve health, wellbeing, and fitness. The message selection system 202 can measure the effectiveness of a rhetorical message by comparing the actions or responses of users who received the rhetorical message to users who did not receive the rhetorical message.
[0155] FIG. 8 illustrates a flow diagram of an example method 800 for indirectly updating a learning model. The method 800 may include identifying, by one or more processors of the server, a plurality of non-rhetorical messages and a plurality of rhetorical messages (block 802). The method 800 may include identifying, by one or more processors of the server, a rhetorical message model (block 804). The method 800 may include identifying, by one or more processors of the server, a plurality of non-rhetorical message models (block 806). The method 800 may include receiving, by one or more processors of the server, a response message to the non-rhetorical message (block 808). The method 800 may include updating the rhetorical message model based on the received response message to the non-rhetorical message (block 810).
[0156] As discussed above, the method 800 may include identifying a plurality of non-rhetorical messages and a plurality of rhetorical messages (block 802). The non-rhetorical messages may include messages that generate a prompt for the user to respond. The rhetorical messages may include messages that do not generate a prompt for the user to respond. For example, the rhetorical messages may include messages that encourage the user, messages that display interesting facts to the user, and the like. Identifying the plurality of non-rhetorical messages and rhetorical messages may include searching a database of messages. The non-rhetorical messages in the database may be tagged or classified by a response classifier. The rhetorical messages in the database may be tagged as rhetorical.
[0157] The method 800 may include identifying a rhetorical message model (block 804). Identifying the rhetorical message model may include identifying message objects that are tagged, grouped, or classified as rhetorical message objects and selecting or identifying a model for the rhetorical message object. For example, a rhetorical message object may be classified as a rhetorical message object having a rhetorical message model. The message selection system may identify that the model for the rhetorical message object needs to be updated. For example, the message selection system may periodically update models for both rhetorical and non-rhetorical message objects. The message selection system may process the message objects daily, weekly, monthly, etc. to select message objects for updating.
[0158] The method 800 may include identifying a plurality of non-rhetorical message models (block 806). The message selection system may identify a plurality of non-rhetorical message objects that have the same classification or grouping as the rhetorical message object (and model) to be updated. For example, the message selection system may determine that a rhetorical message object has a "hydration" tag and a plurality of non-rhetorical message objects tagged with the "hydration" tag. A representation of each message object may be stored in a relational database or other database. An attribute or column of the message object in the database may be a tag or classification tag. A similar message object selector may determine the tag of the rhetorical message object to be updated and query the database for non-rhetorical message objects that have the same tag.
[0159] The method 800 may include receiving a response message to the non-rhetorical message (block 808). The response evaluator may receive the response message to the non-rhetorical message generated by one of the rhetorical message objects identified in block 806. In some implementations, the response evaluator may obtain previous responses to past messages generated by the non-rhetorical message object identified in block 806. The response classifier may generate one or more feature sets based on the received responses or past responses to messages generated by the non-rhetorical message object identified in block 806. For example, the feature set may indicate a time the message was sent to the remote computing device, a time the response was received, and a quality score of the response. The response classifier may also generate a feature set for messages for which the message selection system did not receive a response. The feature set may include a time the message was sent to the remote computing device and an indication that a response was not received.
[0160] The method 800 may include updating the rhetorical message model (block 810). The response evaluator may update the rhetorical model based on a non-rhetorical model of the identified rhetorical message object. For example, in some implementations, the response evaluator may retrain the model of the non-rhetorical message object using feature sets generated in response to responses received (or not received) as training data. The response evaluator may copy the retrained model with the updated feature set to the rhetorical message object being updated. In some implementations, the response evaluator may update the rhetorical message model based on received response messages to the non-rhetorical message. For example, rather than updating the model of the non-rhetorical message object based on responses to that message and copying the updated model to the rhetorical message object, the response evaluator may identify one or more feature sets generated based on responses to messages generated by the identified multiple non-rhetorical message objects. The response evaluator can train a model of the rhetorical message object using one or more feature sets generated based on responses to messages generated by the identified plurality of non-rhetorical message objects.
[0161] D. SYSTEM AND METHOD FOR SELECTING AND SENDING MESSAGES IN A NETWORK ENVIRONMENT USING TRUST VALUES
[0162] Messages may be presented via a computing device to inform a user, such as specifying a particular action to take or providing information. However, when a message is presented repeatedly, the user may not take any action from the message presentation, which may cause message fatigue on the part of the user. Message fatigue may reduce the effectiveness of messages to inform the user and may degrade the quality of human-computer interaction (HCI) of the computing device in presenting the messages. As messages are transmitted over a network and processed by the computing device, the reduced efficiency may lead to wasted computational resources and network bandwidth.
[0163] To address these and other issues related to message fatigue, message selection and delivery may be on a per-user basis. To that end, the message selection system may maintain a set of message objects. Each message may include a template that defines how the corresponding message is generated, and selection criteria that indicate one or more endpoints (or objectives) to be achieved by delivering the corresponding message. When a message is delivered to a particular user, the message selection system may derive or identify the user state (or "intent") using an activity log of the user. The activity log may include entries that indicate how the user responded to previously presented messages. The user state may be used to select one of the message objects to generate the corresponding message.
[0164] The message selection system may generate a confidence value for each message object indicating the likelihood that the presentation of the corresponding message to the user will achieve its purpose. The message selection system may modify the confidence value of the message object based on tuning factors such as the amount of time since any message has been provided to the user's computing device, the total number of messages indicated in the user profile that the computing device is configured to receive within a given time, and the length of time the user is awake or active. With the adjusted confidence value, the message selection system may select one of the message objects. The message selection system may generate and provide a message to the user's computing device according to the selected message object. The user's answers may be recorded using the computing device and provided to the message selection system for use in the next round of selection.
[0165] In this way, the message selection system can send messages that are likely to lead to user actions at the right time, thereby increasing the effectiveness of HCI between the user and the computing device and the presentation of messages on the computing device. In addition, by selecting messages that are likely to lead to user actions, it is possible to reduce the waste of computing resources and network bandwidth required for providing and presenting messages. In addition, by delivering appropriate content tailored to each user at the right time, it is possible to improve the user's engagement with the application. By delivering personalized and relevant messages to each user in a timely manner through the application, it is possible to lead to actual user actions aimed at improving health, well-being, fitness, and quality of life.
[0166] 9A, a block diagram of a system 900 for selecting and sending messages between network environments using trust values is depicted. In overview, the system 900 may include at least one message selection system 902 and one or more remote computing devices 904A-N (collectively referred to hereinafter as remote computing devices 904). The message selection system 902 and the remote computing devices 904 may be communicatively coupled to each other via at least one network 906. The message selection system 902 may include at least one state tracker 908, at least one object evaluator 910, at least one model trainer 912, at least one message distributor 914, at least one message generator 916, at least one response handler 918, at least one evaluation model 920, at least one delivery policy 922, at least one activity database 924, and at least one message database 926. The activity database 924 may maintain or include one or more activity logs 928A-N (hereinafter generally referred to as activity logs 928). The message database 926 may maintain or include one or more message objects 930A-N (hereinafter generally referred to as message objects 930). Each remote computing device 904 may have at least one application 932. The applications 932 running on the remote computing devices 904 may have at least one reporting agent 934. The remote computing devices 904 may be associated with at least one user 936.
[0167] Each component of the system 900 (e.g., the message selection system 902 and its components, and the remote computing device 904 and its components) may be executed, processed, or implemented using hardware or a combination of hardware such as the system 100 detailed herein in Section A. Also, the components of the system 900 may be used to execute, process, or implement functionality detailed herein in Sections B and C. For example, the message selection system 902 may include functionality described in connection with the message selection system 202. The message objects 930 held in the database 924 may include functionality detailed herein in connection with the message objects 206. The object evaluator 912 may include functionality of the message object evaluator 252, among others. The remote computing device 904 may include or execute functionality described herein in connection with the remote computing device 270.
[0168] 9B, a sequence diagram of the system 900 is depicted with the operation of the message selection system 902, specifically the state tracker 908, the model trainer 910, and the object evaluator 912. The state tracker 908 executing on the message selection system 902 may access an activity database 924 to retrieve or identify an activity log 928 of either the remote computing device 904, the application 932, or the user 936. The activity log 928 maintained in the activity database 924 may include one or more entries 940A-N (hereinafter generally referred to as entries 940). Each entry 940 of the activity log 928 may identify or include at least one action by the user 936 recorded via the remote computing device 904 associated with the user 936. The action by the user 936 may be in response to a previous presentation of a message via the application 932 on the remote computing device 904. For example, an application 932 executing on the remote computing device 904 may detect an interaction entered by a user 936 through one or more user interface elements of a previous message presented on the application 932. The interaction with the user interface element may indicate an action performed by the user 936. Once detected, the application 932 may send an indication of the interaction to the message selection system 902 for recording in an activity log 928 on the activity database 924. The indication may include a timestamp identifying the time the interaction was entered. Once received, the indication from the remote computing device 904 may be stored and maintained in the activity database 924 as one of the entries 940 in the activity log 928.
[0169] Upon identification from the activity database 924, the state tracker 908 may parse the activity log 928 to identify or retrieve entries 940. By identifying the entries 940, the state tracker 908 may determine, identify, or retrieve actions performed by the user 936 recorded via the remote computing device 904, as indicated in the entries 940 of the activity log 928. In some embodiments, the state tracker 908 may select or identify a subset of the entries 940 of the activity log 928 based on the timestamps of the entries 940 within a defined time window. The defined time window may be relative to the current time and may range from one hour to one month. For example, the state tracker 908 may select a subset of the entries 940 of the activity log 928 from the past week, as indicated by the timestamps. Upon selection form the activity log 928, the state tracker 908 may identify the actions performed by the user 936, as indicated in each entry 940.
[0170] Based on the actions indicated in the entries 940, the state tracker 908 may identify or determine at least one user state 942 (sometimes referred to herein as “intent” or “user intent”) for the user 936 of the remote computing device 904. The user state 942 may infer the actions recorded in the entries 940 of the activity log 928 on the activity database 924 to classify or identify an expected goal (e.g., a behavioral endpoint) on the part of the user 936. The user state 942 may relate to various aspects of the usage actions indicated in the entries 940, and may include, for example, application navigation events (e.g., opening the application 932), engagement patterns or interaction rates with one or more features such as completion of the application 932 belonging to a particular milestone in compliance with a regime provided via the application 932, and user input data from a survey (e.g., related to the user's health), etc. The user state 942 may be used as one of the parameters in selecting one of the message objects 930 corresponding to a message to provide for presentation to the user 936. In some embodiments, the user state 942 may correspond to context data 220 as discussed in sections B and C, and the state tracker 908 may select or determine a user state 942 from a set of user states 942 based on the action identified from the activity log 928. Each user state 942 in the set may be defined as corresponding to one or more actions. For example, if the action identified from the activity log 928 matches one of the actions defined for the user states 942 in the set, the state tracker 908 may determine the corresponding user state 942 for the user 936.
[0171] Relatedly, an object evaluator 910 executing on the message selection system 902 may identify one or more message objects 930 from the message database 926. The identified message objects 930 may be used to select and generate corresponding messages for presentation at the remote computing device 904. candidate Each message object 930 may include a set of executable instructions (e.g., a script) for evaluating whether to generate and send a corresponding message for presentation to a user 936. In some embodiments, message objects 930 may correspond to message objects 206 and their components described above in sections B and C.
[0172] Each message object 930 may include, among other things, at least one message template 944, at least one constraint 946, and at least one selection criterion 948. The message template 944 may include instructions for generating a corresponding message to include one or more user interface elements. For example, the message template 944 may identify placeholders for inserting text or visual content into graphical user interface elements of a message to be presented via the remote computing device 904. The constraints 946 may define or include one or more constraints for generating or sending a message corresponding to the message object 930. The constraints 946 may include, for example, one or more time windows (e.g., time of day or weekends vs. weekdays) during which the generation and sending of the corresponding message is permitted. The selection criterion 948 may define or include one or more endpoints that the user 936 should achieve when submitting a message corresponding to the message object 930. The selection criterion 948 may also identify or define a time window within which the endpoints should be achieved. The endpoints of the selection criteria 948 may include behavioral goals that are expected to be achieved by the user 936, such as motivating the user 936 to engage in a regimen pattern to enable optimal delivery of the digital therapy provided via the application 932. The selection criteria 948 may be pre-configured or pre-set with the goal of motivating the user 936 to receive greater therapeutic benefits. The behavioral objects defined by the selection criteria 948 may include, for example, application navigation events (e.g., opening the application 932), performance of a particular activity (e.g., breathing, exercising, or dieting) on the part of the user 936, and scheduling an activity via the application 932 (e.g., specifying a future time to perform an exercise), etc.
[0173] For each message object 930, the object evaluator 910 may calculate, generate, or determine at least one confidence value 952 based on the user state 942 and the selection criteria 948. The confidence value 952 may identify, represent, or indicate a predicted effect of a message corresponding to the message object 930 on a particular user 936 toward achieving an endpoint defined by the selection criteria 948 for the message object 930. The confidence value 952 may also indicate a predicted likelihood that the user 936 will take an action corresponding to at least one of the endpoints defined by the selection criteria 948. In determining the confidence value 952 for the message object 930, the object evaluator 910 may compare the endpoints defined by the selection criteria 948 with the user state 942 of the user 938. The user state 942 may identify an intent on the part of the user 936 and may correspond to one of the endpoints defined by the selection criteria 948. If at least one of the endpoints defined by the selection criteria 948 matches or corresponds to the user state 942, the object evaluator 910 may determine a relatively high confidence value 942. Conversely, if none of the endpoints defined by the selection criteria 948 matches or corresponds to the user state 942, the object evaluator 910 may determine a relatively low confidence value 942. In some embodiments, the message distributor 914 may also remove the message object 930 from further processing.
[0174] To facilitate the determination of the confidence values 952, the model trainer 912 executing on the message selection system 902 may maintain an evaluation model 920. The evaluation model 920 may be used to determine the confidence values 952 of the message objects 930. The evaluation model 920 may include, for example, a regression model (e.g., a linear regression model or a logistic regression model), a support vector machine (SVM), an artificial neural network (ANN), a naive Bayes classifier, a classification model (e.g., k-nearest neighbors), and the like. In general, the evaluation model 920 may include a set of inputs and a set of outputs related to each other via a set of weights. In some embodiments, the evaluation model 920 may be part of or dedicated to one of the message objects 930 (e.g., model 210). In some embodiments, the evaluation model 920 may be general to the set of message objects 930 held in the message database 926, separate from the individual message objects 930.
[0175] The model trainer 912 may use the training dataset to initiate or establish the evaluation model 920. Training of the evaluation model 920 may be performed depending on the architecture of the evaluation model 920. For example, if the evaluation model 920 is a classification model, training may be performed iteratively until convergence. The training dataset for the evaluation model 920 may include historical response data from users (e.g., users 936) to previously presented messages. The messages in the training dataset may correspond to one or more of the message objects 930. The training dataset may include message templates 944, constraints 946, and selection criteria 948 of the message objects 930 that correspond to messages previously provided to the users. The historical response data may indicate success or failure in achieving the endpoints specified by the message objects 930. In training the model, the model trainer 912 may apply the training dataset to the evaluation model 920 to generate an output indicating success or failure. The model trainer 912 may adjust or modify weights of the evaluation model 920 based on a comparison of the output to the results indicated in the historical response data, and repeat the application until convergence. Modification of the weights of the evaluation model 920 may follow learnings on the architecture used to implement the evaluation model 920 .
[0176] Upon establishment, the object evaluator 910 may apply the user state 942 and the message object 930 to the evaluation model 920 to generate or determine a confidence value 952. In some embodiments, the object evaluator 910 may consider the message template 944, the constraints 946, and the selection criteria 948, or any combination thereof (e.g., in their feature space representations) when applying to the evaluation model 920. For example, the object evaluator 910 may use the endpoints defined by the user state 942 and the selection criteria 948 of the message object 930 when applying to the evaluation model 920. To apply, the object evaluator 910 may provide the user state 942 and the message object 930 (e.g., in their feature space representations) as inputs to the evaluation model 920. The object evaluator 910 may process the inputs through the use of the evaluation model 920 weights to generate an output of the evaluation model 920. The object evaluator 910 may identify the output of the evaluation model 920 as a confidence value 952 for the message object 930. In general, when at least one of the endpoints defined by the selection criteria 948 correlates with the user state 942, the confidence value 942 may be relatively high. Conversely, when none of the endpoints defined by the selection criteria 948 correlates with the user state 942, the confidence value 942 may be relatively low.
[0177] In some embodiments, the message distributor 914 may keep or remove one or more of the message objects 930 from the selection based on the constraints 946 defined for the message objects 930. For each message object 930, the message distributor 914 may identify a constraint 946. As described above, the constraints 946 may define or include one or more constraints (e.g., a time window) for generating or sending a message corresponding to the message object 930. In some embodiments, the message distributor 914 may identify a current time that corresponds to a time at which a message corresponding to the message object 930 will be sent. The message distributor 914 may compare the current time to the restricted time window defined by the constraints 946 of the message object 930. When the current time is not within the restricted time window defined by the constraints 946, the message distributor 914 may keep the message object 930 for further processing, as described below. Otherwise, when the current time is within the restricted time window defined by the constraints 946, the message distributor 914 may remove the message object 930 from further processing.
[0178] 9C, a sequence diagram of a system 900 having operations of a message distributor 914 and a message generator 916, among other message selection system 902, is depicted. A message distributor 914 executing on the message selection system 902 may identify or select at least one of the message objects 930 according to at least one delivery policy 922. The delivery policy 922 may specify the selection of at least one of the message objects 930 based on, among other factors, a trust value 952 for the message object 930. In some embodiments, the delivery policy 922 may be specific to one of the message objects 930. In some embodiments, the delivery policy 922 may be applied generally to the set of message objects 930.
[0179] The delivery policy 922 may specify, define, or otherwise include, among other things, at least one update factor 960, at least one threshold 962, and at least one delivery time 964. The update factor 960 may specify or define a value or amount by which the trust value 952 for the message object 930 should be modified or adjusted. In some embodiments, the update factor 960 may specify that the amount by which the trust value 952 is adjusted (e.g., increased or decreased) is based on the amount of time elapsed since the last provision of any message to the remote computing device 904. For example, the update factor 960 may function as a cool-down policy 232, as described in sections B and C. The threshold 962 may specify a value of the trust value 952 at which the message object 930 is permitted to be selected for further processing. In some embodiments, the threshold 962 may be defined relative to the trust value 952 determined for the message object 930 maintained on the message database 926. For example, threshold 962 may specify that message objects 930 having up to 5% of trust value 952 are permitted to be selected for further processing. In some embodiments, threshold 962 may function as threshold policy 234 as discussed in Sections B and C. Delivery time 964 may specify a time or window of time during which a message corresponding to one of the message objects 930 may be sent or provided to a remote computing device 904.
[0180] In determining which message object 930 to select, the message distributor 914 may use an update factor 960 of the delivery policy 922 to modify or otherwise update the trust value 952 of each message object 930. In some embodiments, the message distributor 914 may determine an amount of time that has elapsed between the last time that any message corresponding to one of the message objects 930 was sent to the remote computing device 904 and the current time. The determination of the amount of time that has elapsed may be determined by the update factor 960 of the delivery policy 922. Based on the amount of time that has elapsed, the message distributor 914 may determine an update factor 960 for the message object 930. Using the update factor 960, the message distributor 914 may modify or otherwise update the trust value 952 of each message object 930. For example, the message distributor 914 may multiply the trust value 952 by the update factor 960 to adjust the trust value 952.
[0181] In some embodiments, in addition to the specification of the delivery policy 922, the message distributor 914 may keep or remove one or more of the message objects 930 from the selection based on the constraints 946 defined for the message objects 930. For each message object 930, the message distributor 914 may identify a constraint 946. As described above, the constraints 946 may define or include one or more constraints (e.g., a time window) for generating or sending a message corresponding to the message object 930. In some embodiments, the message distributor 914 may identify a current time that corresponds to a time at which a message corresponding to the message object 930 will be sent. In some embodiments, the message distributor 914 may compare the current time to the restricted time window defined by the constraints 946 of the message object 930. When the current time is not within the restricted time window defined by the constraints 946, the message distributor 914 may keep the message object 930 for further processing, as described below. Otherwise, if the current time is within a restricted time window defined by constraints 946, the message distributor 914 may remove the message object 930 from further processing.
[0182] Based on the updated trust value 952, the message distributor 914 may identify or select at least one of the set of message objects 930. In some embodiments, the message distributor 914 may rank the set of message objects 930 by the updated trust value 952. Using the ranking, the message distributor 914 may select one of the message objects 930 (e.g., the message object 930 having the highest trust value 952 in the ranking). In some embodiments, the message distributor 914 may compare the updated trust value 952 of the message object 930 to a threshold 962 defined in the delivery policy 922. If the trust value 952 meets (e.g., is greater than) the threshold 962, the message distributor 914 may select the corresponding message object 930. Conversely, if the trust value 952 does not meet (e.g., is less than) the threshold 962, the message distributor 914 may refrain from selecting the corresponding message object 930.
[0183] Upon selection of the message object 930, a message generator 916 executing on the remote computing system may create or generate at least one corresponding message 916 for presentation to a user 936 of the remote computing device 904. The message generator 916 may identify a message template 944 for the selected message object 930. Upon identification, the message generator 916 may parse the message template 944 specification for the message object 930. Upon parsing, the message generator 916 may execute the message template 944 to generate a message 966. Once generated, the message 966 may include one or more user interface elements 968A-N (collectively referred to hereinafter as user interface elements 968). The user interface elements 968 may correspond to graphical user interface elements of the message 966 presented via the remote computing device 904 and may include functionality that is activated in response to detection of an interaction by the user 936.
[0184] 9D, a sequence diagram of the system 900 having the operation of the message generator 916 and the response handler 918 of the message selection system 902 and the remote computing device 914 is depicted. Upon generation, the message generator 916 may provide, transmit, or otherwise send a message 966 to the remote computing device 904 over the network 906. In some embodiments, the message deliverer 914 may compare the current time to a specified time window defined by a delivery time 964 of the delivery policy 922. When the current time is within the specified time window defined by the delivery policy 922, the message deliverer 914 may send the message object 930 to the remote computing device 904. Otherwise, when the current time is outside the specified time window defined by the delivery policy 922, the message deliverer 914 may wait until the current time matches the delivery time 964.
[0185] Upon receipt, an application 932 executing on the remote computing device 904 may present the message 966 for the user 936. The application 932 may configure the remote computing device 904 to communicate with the message selection system 902 over the network 906. The application 932 may parse the message 966 and present (e.g., render or play) the message 966, including the user interface element 968, via the remote computing device 904. Once the message 966 is presented, a reporting agent 934 of the application 932 may monitor one or more interactions with the user interface element 968 of the message 966 via an input / output (I / O) device of the remote computing device 904. The interactions may be performed by the user 936 and may be used to record one or more activities performed by the user 936 in response to the presentation of the message 966. For example, if the message 966 presented via the remote computing device 904 indicates that the user 936 is to perform a breathing exercise, the user 936 may select a command button that indicates that the breathing exercise has been performed. Based on the interactions detected with the message 966, the reporting agent 934 may generate and transmit at least one response 970. The response 970 may identify one or more actions taken by the user 936 upon presentation of the message 966 that were recorded using interactions via the remote computing device 904. The response 970 may also identify or include a timestamp corresponding to the time the action was recorded on the remote computing device 904. Upon generation, the application 932 may send or transmit the response 970 to the message selection system 902 over the network 906.
[0186] A response handler 918 executing on the message selection system 902 may in turn identify or receive a response 970 from the remote computing device 904 over the network 906. The response handler 918 may parse the response 970 to identify one or more actions taken by the user 936 as indicated in the response 970. Using the activities identified from the response 970, the response handler 918 may generate one or more entries 940′A-N (collectively referred to hereinafter as entries 940′) for inclusion in an activity log 928 of the user 936. The entries 940 may identify actions taken by the user 936 in response to presentation of the message 966. The entries 940 may include a timestamp corresponding to the time the user 936 recorded the action on the application 932. The response handler 918 may append or insert the new entries 940′ into the activity log 928 of the user 936 maintained in the message database 924. By adding entry 940′, the response handler 918 may record the actions taken by the user 936 in response to the presentation of the message 966 via the remote computing device 904. The newly added entry 940′ to the activity log 928 may be used to update the user state 942 for the user 936.
[0187] In this manner, the message selection system 902 may select and generate messages 966 that are more likely to be interacted with by a user 936 associated with the remote computing device 906 at the appropriate time. The selection and generation of messages 966 may enhance the quality of human-computer interaction (HCI) between the user 936 and the system 900 as a whole (including the remote computing device 904). The selection and generation may also result in an increased likelihood of a behavioral endpoint or goal for the user 936. Furthermore, because the messages 966 are more likely to be read and interacted with, the message selection system 902 may better utilize and reduce wasted computing resources and network bandwidth.
[0188]
[0031] Referring now to FIG. 10, there is depicted a flow diagram of a method 1000 for selecting and sending a message in a network environment using a trust value. The method 1000 may be implemented using any of the components as detailed herein in connection with FIGS. 1-9D. 1-9D. In brief overview, a server may obtain actions from an activity log (1005). The server may determine a user state (1010). The server may identify message object candidates (1015). The server may determine a trust value (1020). The server may update the trust value with time (1025). The server may select a message object (1030). The server may generate a message (1035). The server may send the message (1040).
[0189] In more detail, a server (e.g., message selection system 902) may retrieve (1005) the actions from an activity log (e.g., activity log 928). The activity log may include one or more entries (e.g., entries 940). Each entry in the activity log may identify one or more actions performed by a user (e.g., user 936) in response to a previous message submission. The server may parse the activity log entries to identify the one or more actions. The server may determine (1010) a user state (e.g., user state 942). Using the identified actions, the server may determine a user state for the user. The user may indicate the user's intent in performing the action recorded in the activity log. The server may also select a user state from a set of defined user states. Each user state may correspond to one or more actions performed by the user.
[0190] The server may identify (1015) candidate message objects (e.g., message objects 930). Each message object may include, among other things, a message template (e.g., message template 944), constraints (e.g., constraints 946), and selection criteria (e.g., selection criteria 948). The message template may specify the generation of a corresponding message and user interface elements on the message. The constraints may specify conditions that limit the selection of the message object. The selection criteria may define one or more endpoints that should be achieved when submitting a message corresponding to the message object.
[0191] The server may determine (1020) a trust value (e.g., trust value 952). To determine a trust value for each message object, the server may compare the user state to the endpoints defined by the selection criteria of the message object. The trust value may indicate a predicted effect of the corresponding message on the user toward achieving the endpoints identified by the selection criteria of the message object. The server may also apply the message object and the user state to a reputation model (e.g., reputation model 920) to generate the trust value. The server may update (1025) the trust value with elapsed time. The server may modify the trust value based on the amount of time that has elapsed since the last time the message was presented to the user. The server may increase or decrease the trust value based on the amount of time that has elapsed.
[0192] The server may select a message object (1030). The server may select one of the message objects based on the trust value. The server may compare the trust value of the message object to a threshold (e.g., threshold 962). If the trust value meets the threshold, the server may select the message object. Otherwise, when the trust value does not meet the threshold, the server may refrain from selecting the message object. The server may generate a message (1035) (e.g., message 966). Using a template of the selected message object, the server may generate the message. The message may include one or more user interface elements (e.g., user interface element 968). The server may send the message (1040). The message may be sent to a remote computing device (e.g., remote computing device 904). Upon receipt, the remote computing device may present the message.
[0193] E. SYSTEM AND METHOD FOR SELECTING AND SENDING DIFFERENT TYPES OF MESSAGES OVER A NETWORK ENVIRONMENT
[0194] However, repeated presentation of a message may cause message fatigue on the part of the user, leading to the user taking no action from the presentation of the message and leading to a decrease in the effectiveness of the message to inform the user. This may lead to a decrease in the quality of the human-computer interaction (HCI) of the computing device in presenting the message. The decrease in effectiveness may also lead to a waste of computational resources and network bandwidth as the message is transmitted over a network and processed by the computing device. This may be particularly problematic when evaluating the performance of messages that contain information for a user but are not configured to receive input via a user interface element (sometimes referred to herein as a rhetorical message). For messages that are configured to receive input via a user interface element (sometimes referred to herein as a non-rhetorical message), the performance score of such messages may be determined based on recorded responses. In contrast, for messages that are not configured to receive input, there may be no recorded responses to the message.
[0195] To address these and other technical challenges in assessing the effectiveness of messages, the performance scores of messages with recorded responses may be used to estimate the predicted performance scores of user interface elements lacking messages. The message selection system may identify a set of message objects for which responses to a corresponding message have been recorded. Each message object may define an endpoint to be achieved on the part of a user in presenting the corresponding message. The message selection system may compare the responses from the users with the endpoints of the message objects. From the comparison, the message selection system may determine the number of users who satisfied the endpoints of the message objects corresponding to the messages presented to the users.
[0196] Based on the number of users, the message selection system may determine a performance score for each message object. The performance score may indicate a degree of effectiveness of presenting a message to the user. Upon this determination, the message selection system may rank the message objects by their performance scores. Using the ranking, the message selection system may select a set of message objects to allow for use in generating a message for transmission. The message selection system may also select another set of message objects to restrict for use in generating a message. From the set of allowed message objects, the message selection system may identify other message objects that lack the ability to record a response due to configuration without an input element. The identification of the other message objects may be based on corresponding endpoints. The message selection system may add the identified message objects to the set of allowed message objects.
[0197] In this manner, the message selection system may expand the set of messages that can be generated and provided to a remote computing device, thereby enhancing the functionality of the messages. Message expansion can improve the quality of human-computer interaction (HCI) between a user and a remote computing device. Also, because messages are more likely to result in user action (even if not recorded), the message selection system can make better use of and reduce wasted computing resources and network bandwidth consumed in providing and presenting messages.
[0198] 11A, a block diagram of a system 1100 for selecting and sending different types of messages across a network environment using confidence values is depicted. In overview, the system 1100 may include at least one message selection system 1102 and one or more remote computing devices 1104A-N (collectively referred to hereinafter as remote computing devices 1104). The message selection system 1102 and the remote computing devices 1104 may be communicatively coupled to each other via at least one network 1106. The message selection system 1102 may include at least one response tracker 1108, at least one object evaluator 1110, at least one model trainer 1112, at least one delivery controller 1114, at least one message correlator 1116, at least one performance model 1118, at least one response database 1120, and at least one message database 1122, etc. The response database 1120 may maintain or include response data 1124A-N (collectively referred to hereinafter as response data 1124). The message database 1122 may maintain or include a set of message objects 1126A-N of one type (collectively referred to hereinafter as message objects 1126) and a set of message objects 1126′A-N of another type (collectively referred to hereinafter as message objects 1126′). Each remote computing device 1104 may include at least one application 1128. The applications 1128 running on the remote computing device 1104 may include at least one reporting agent 1130. The remote computing device 1104 may be associated with at least one user 1132.
[0199] Each component of the system 1100 (e.g., the message selection system 1102 and its components, and the remote computing device 1104 and its components) may be executed, processed, or implemented using hardware or a combination of hardware such as the system 100 detailed herein in Section A. Also, the components of the system 1100 may be used to execute, process, or implement functionality detailed herein in Sections B and C. For example, the message selection system 1102 may include functionality described in connection with the message selection system 202. The message objects 1126 or 1126′ held in the database 1120 may include functionality detailed herein in connection with the message objects 206. The object evaluator 1110 may include functionality of the message object evaluator 252, etc. The remote computing device 1104, in conjunction with the remote computing device 270, may include or execute functionality described herein.
[0200] 11B, a sequence diagram of the system 1100 having the operation of a response tracker 1108, an object evaluator 1110, and a model trainer 1112 of the message selection system 1102 is depicted. The response tracker 1108 executing on the message selection system 1102 may access a response database 1120 to obtain or identify response data 1112. The response data 1112 may include one or more entries 1140A-N (collectively referred to hereinafter as entries 1140) spanning one or more of the remote computing device 1104, the application 1126, or the user 1130. In some embodiments, the response tracker 1108 may identify a subset of the entries 1140 from the response data 1124 within a time window. The time window may be defined relative to the current time and may be between one hour and one month. Response data 1124 collected from a time window may be used to evaluate or assess the performance of messages corresponding to message objects 1126 previously presented to a user 1130 .
[0201] Each entry 1140 of the response data 1120 (sometimes referred to herein as a response) may identify or include at least one action by the user 1132 recorded via a remote computing device 1104 associated with the user 1132. The action by the user 1132 may be in response to a previous presentation of a message via an application 1128 on the remote computing device 1104. For example, the application 1128 running on the remote computing device 1104 may detect an interaction entered by the user 1132 via one or more of the user interface elements of a previous message presented on the application 1128. The previous message may have been generated using a first type of message object 1126. The interaction with the user interface element may indicate an action performed by the user 1132. Once detected, the application 1128 may send an indication of the interaction to the message selection system 1102 for recording in the response data 1124 on the response database 1120. The indication may include a timestamp identifying the time the interaction was entered. Upon receipt, the instruction from the remote computing device 1104 may be stored and maintained on the response database 1120 as one of the entries 1140 of the response data 1124 .
[0202] In some embodiments, the response tracker 1108 may use the entries 1140 of the response data 1112 of the users 1132 to identify or determine a user state (sometimes referred to herein as “intent” or “user intent”) for each user 1132. The determination of the user state by the response tracker 1108 is performed in a manner similar to the determination of the user state 942 by the state tracker 908 described above. The user state may infer the actions recorded in the entries 1140 of the response data 1124 to classify or identify an expected goal (e.g., a behavioral endpoint) on the part of the user 1132. The user state may relate to various aspects of the usage activity depicted in the entries 1140. The entries 1140 may, for example, be application navigation events (e.g., opening the application 1128), engagement patterns or interaction rates with one or more features such as completion of the application 932 belonging to a particular milestone in compliance with a regime provided via the application 932, and user input data from a survey (e.g., related to the user's health), etc. In some embodiments, the response tracker 1108 may determine a user state for the user 1132 prior to presentation of the message using a subset of the pre-presentation entries 1140. The subset of entries 1140 from the response data 1124 may include, for example, those used in selecting the message or a set number of entries 1140 prior to presentation of the message. The response tracker 1108 may identify the pre-presentation entries 1140 based on a timestamp of the entries 1140. In some embodiments, the response tracker 1108 may select or determine a user state from a set of user states based on actions identified from the response data 1124. Each user state in the set may be defined as corresponding to one or more actions.
[0203] An object evaluator 1110 executing on the message selection system 1102 may identify one or more message objects 1126 of a first type (sometimes referred to herein as a non-rhetorical type) from the message database 1122. Each message object 1126 may include a set of executable instructions (e.g., a script) for evaluating whether to generate and send a corresponding message 1146 for presentation to a user 1132. In some embodiments, the message objects 1126 may correspond to the message objects 206 and their components described above in sections B and C. In some embodiments, the object evaluator 1110 may identify a message object 1126 of the first type.
[0204] Each message object 1126 may include at least one message template 1142 and at least one selection criterion 1144, etc. The message template 1142 may include instructions for generating at least one corresponding message 1146 of a first type to include one or more user interface elements. For example, the message template 1142 may identify placeholders for inserting text or visual content into graphical user interface elements of a message to be presented via the remote computing device 1104. The selection criterion 1144 may define or include one or more endpoints to be achieved by the user 1132 when submitting the message 1146 corresponding to the message object 1126. The selection criterion 1144 may also identify or define a time window within which the endpoints should be achieved. The endpoints of the selection criterion 1144 may include behavioral goals expected to be achieved on the part of the user 1132, such as, for example, as described above with respect to the behavioral goals defined by the selection criterion 948. The selection criterion 1144 may also indicate which actions should be performed by the user 1132 to satisfy one or more of the endpoints.
[0205] The message template 1142 of each message object 1126 of the first type may correspond to or may be used to generate at least one message 1146. For a message object 1126 of the first type, each message 1146 may include one or more non-interactive elements 1148A-N (collectively, hereinafter, non-interactive elements 1148) and one or more interactive elements 1150A-N (collectively, hereinafter, interactive elements 1150). Each non-interactive element 1148 may lack a configuration for responding to a detected interaction by a user 1132 when presented via a remote computing device 1104. Also, the non-interactive element 1148 may lack a configuration for recording an interaction by a user 1132. In contrast, each interactive element 1150 may have a configuration for responding to a detected interaction by a user 1132 when presented via a remote computing device 1104. The interactive element 1150 may also be configured to record interactions by the user 1132 and communicate the detected interactions with the message selection system 1102 .
[0206] For each message object 1126, the object evaluator 1110 may identify response data 1124 that records actions taken by the user 1130 when a message 1146 corresponding to the message object 1126 was presented. In identifying the response data, the object evaluator 1110 may access the response database 1120 to identify one or more of the entries 1140 of the response data 1124 that correspond to a response to presentation of the corresponding message 1146. In each identified entry 1140, the object evaluator 1110 may identify input by the user 1132 to one or more of the interactive elements 1152 of the message 1146. The input may be recorded via an interaction with one of the interactive elements 1152 of the message 1146. The input may also be indicative of an action taken by the user 1132 recorded via the remote computing device 1104 when the corresponding message 1146 was presented. In some embodiments, the object evaluator 1110 may cooperate with the response tracker 1108 to analyze the entries 1140 to identify recorded actions performed by the user 1132 .
[0207] Based on input from each entry 1140 of the message object 1126, the object evaluator 1110 may determine whether at least one of the endpoints defined by the selection criteria 1144 of the message object 1126 has been satisfied. To determine, the object evaluator 1110 may compare the actions indicated in the entries 1140 to the endpoints defined by the selection criteria 1144 of the message object 1126. As described above, the selection criteria 1144 of the message object 1126 may specify one or more actions for the user 1132 to perform to satisfy the defined endpoints. In some embodiments, the object evaluator 1110 may traverse the entries 1140 of the message object 1126 when comparing the actions to the endpoints.
[0208] When the action identified from the entry 1140 matches one of the actions specified by the endpoints of the selection criteria 1144, the object evaluator 1110 may determine that at least one of the endpoints of the message object 1126 is satisfied. From that determination, the object evaluator 1110 may count or determine the number of users 1130 for each message object 1126 that satisfied at least one of the endpoints specified in the selection criteria 1144 of the message object 1126. Conversely, when the action identified from the entry 1140 does not match any of the actions specified by the endpoints of the selection criteria 1144, the object evaluator 1110 may determine that none of the endpoints of the message object 1126 are satisfied. From this determination, the object evaluator 1110 may count or determine the number of users 1130 for each message object 1126 that did not satisfy any of the endpoints specified by the selection criteria 1144 of the message object 1126.
[0209] Using the number of users 1130 that satisfied at least one of the endpoints defined by the selection criteria 1144 of the message object 1126, the object evaluator 1110 may calculate, generate, or determine at least one performance score 1154 for the message object 1126. In some embodiments, in determining the performance score 1154, the number of users 1130 that did not satisfy any endpoints defined by the selection criteria 1144 of the message object is calculated. The performance score 1152 may identify, represent, or indicate the effectiveness of the corresponding message 1146 in achieving the endpoints defined by the selection criteria 1144. The performance score 1152 may also indicate the percentage of users 1130 that satisfied one or more endpoints defined by the selection criteria 1144 of the message object 1126. In some embodiments, the object evaluator 1110 may determine the performance score 1152 using a function (e.g., a weighted average). The function may take into account, among other factors, the number of users 1130 who satisfied at least one of the endpoints and the number of users 1130 who did not satisfy any of the endpoints. In general, if a large number of users are determined to have satisfied at least one of the endpoints defined by the selection criteria 1144 of the message object 1142, the performance score 1152 may be relatively high. Conversely, if a small number of users are determined to have satisfied at least one of the endpoints defined by the selection criteria 1144 of the message object 1142, the performance score 1152 may be relatively low.
[0210] In some embodiments, in determining the performance score 1152, the object evaluator 1110 may compare a user state of the user 1132 associated with the entry 1140 prior to submission of the corresponding message 1146 to the endpoints defined for the corresponding message object 1126. When the user state matches or corresponds to at least one of the endpoints defined in the selection criteria 1144 for the message object 1126, the object evaluator 1110 may determine a relatively high performance score 1152. Conversely, when the user state does not match or correspond to any of the endpoints defined in the selection criteria 1144 for the message object 1126, the object evaluator 1110 may determine a relatively low performance score 1152.
[0211] To facilitate the determination of the performance scores 1152, the model trainer 1112 running on the message selection system 1102 may maintain a performance model 1118. The performance model 1118 may be used to determine the performance scores 1152 of the message objects 1126. The performance model 1118 may include, for example, a regression model (e.g., a linear regression model or a logistic regression model), a support vector machine (SVM), an artificial neural network (ANN), a naive Bayes classifier, and a classification model (e.g., k-nearest neighbors), etc. In general, the performance model 1118 may include a set of inputs and a set of outputs related to each other via a set of weights. In some embodiments, the performance model 1118 may be dedicated to a portion or one of the message objects 1126 (e.g., model 210). In some embodiments, the performance model 1118 may be separate from the individual message objects 1126 and may be general to the set of message objects 1126 (and message objects 1126').
[0212] The model trainer 1112 may use a training dataset to initiate or establish a performance model 1118. Training of the performance model 1118 may be performed depending on the architecture of the performance model 1118. For example, if the performance model 1118 is a classification model, it may be iteratively trained until convergence. The training dataset for the performance model 1118 may include historical response data from a user (e.g., user 1132) to previously presented messages. The messages in the training dataset may correspond to one or more of the message objects 1130. The training dataset may include message templates 1142 and selection criteria 1144 of the message objects 1130 that correspond to messages previously provided to the user. The historical response data may indicate success or failure in achieving an endpoint specified by the message object 1130. In training the model, the model trainer 1112 may apply the training dataset to the performance model 1118 to generate an output indicating success or failure. The model trainer 1112 may adjust or modify the weights of the performance model 1118 based on a comparison of the outputs to the results indicated by the historical response data, and may iterate until convergence. Modification of the weights of the performance model 1118 may follow learnings about the architecture used to implement the performance model 1118.
[0213] In some embodiments, the model trainer 1112 may use at least a portion of the response data 1124 to modify or update the performance model 1118. The application of the response data 1124 may occur prior to, in conjunction with, or subsequent to the determination of the performance score 1152. Training the performance model 1118 with the response data 1124. In some embodiments, at least a portion of the response data 1124 may be included as part of a training data set. In some embodiments, the model trainer 1112 may use the entries 1140 of the response data 1124 to update the performance model 1118 based on a determination of whether an endpoint defined by the message object 1126 is met. For example, in training, the model trainer 1112 may apply the entries 1140 to the performance model 1118 to generate an output indicative of success or failure. The model trainer 1112 may adjust or modify weights of the performance model 1118 based on a comparison of the output to a determination of whether the endpoint is met, and repeat the application until convergence. Modification of the weights of the performance model 1118 may follow learnings on the architecture used to implement the performance model 1118 .
[0214] With the establishment of the performance model 1118, the object evaluator 1110 may apply the response data 1124 identified for the message object 1126 to determine a performance score 1152. In some embodiments, the object evaluator 1110 may also consider the message template 1142 and the selection criteria 1144 of the message object 1126 in determining the performance score 1152. In some embodiments, the object evaluator 1110 may also consider a user state determined from the entries 1140 of the message object 1126. To apply, the object evaluator 1110 may provide the entries 1140 of the response data 1124 identified for the message object 1126 and the message object 1126 itself as inputs to the performance model 1118. The object evaluator 1110 may process the inputs through the use of the weights of the performance model 1118 to generate an output of the performance model 1118. The object evaluator 1110 may determine the output of the performance model 1118 as a performance score 1152 for the message object 1126. In general, if a large number of users are determined to have satisfied at least one of the endpoints defined by the selection criteria 1144 of the message object 1142, the performance score 1152 may be relatively high. Conversely, if a small number of users are determined to have satisfied at least one of the endpoints defined by the selection criteria 1144 of the message object 1142, the performance score 1152 may be relatively low.
[0215] In some embodiments, the object evaluator 1110 may assign or attribute one or more additional endpoints to the selection criteria 1144 of at least one of the message objects 1126 based on the response data 1124 for the message object 1126. The object evaluator 1110 may identify a user state from the entries 1140 prior to the presentation of the message 1146 corresponding to the message object 1126. Based on the identification of the user state, the object evaluator 1110 may identify an endpoint corresponding to the user state. In some embodiments, the object evaluator 1110 may select an endpoint from a set of endpoints defined with the user state. The set may define a mapping between the user state and one or more endpoints for inclusion in the selection criteria 1144. The mapping may be defined in the message object 1126 or the message database 1122. Upon finding a match between the user state and one or more endpoints, the object evaluator 1110 may add or insert the endpoint into the definition of the selection criteria 1144 of the message object 1126. In some embodiments, the object evaluator 1110 may add an identified endpoint if the performance score 1152 of the message object 1126 meets (e.g., is greater than or equal to) a threshold score. The threshold score may define the value of the performance score 1152 when adding the endpoint to the selection criteria 1144 of the message object 1126.
[0216] 11C, a sequence diagram of the system 1100 is depicted, including the operation of a delivery controller 1114 within the message selection system 1102. The delivery controller 1114 executing on the message selection system 1102 can determine whether to assign or include each message object 1126 in a permitted set 1160 or a restricted set 1162 based on a corresponding performance score 1152. Both the permitted set 1160 and the restricted set 1162 may be maintained on the message selection system 1102 (e.g., in the message database 1122). Message objects 1126 assigned to the permitted set 1160 may be permitted to be selected for generation and provision to the remote computing device 1104. In some embodiments, message objects 1126 assigned to the permitted list 1160 may be more likely (e.g., more highly weighted) to be selected for generation and provision for presentation via the remote computing device 1104. On the other hand, message objects 1126 assigned to the restricted set 1162 may be limited or prevented from being generated and selected for transmission to a remote computing device 1104. In some embodiments, message objects 1126 assigned to the restricted set 1162 may be made less likely (e.g., weighted lower) for selection to be generated and provided for presentation to a remote computing device 1104.
[0217] To determine the assignment, the delivery controller 1114 may rank the message objects 1126 based on their performance scores 1152. The ranking of the message objects 1126 may be according to a sorting algorithm such as insertion sort, merge sort, heap sort, quick sort, bubble sort, radix sort, or bucket sort, among others. Based on the ranking, the delivery controller 1114 may assign the message objects 1126 to the allowed set 1160 or the restricted set 1162. In some embodiments, the delivery controller 1114 may select a subset of the message objects 1126 with the highest n performance scores 1152 to assign to the allowed set 1160 and assign the remaining message objects 1126 to the restricted set 1162. In some embodiments, conversely, the delivery controller 11144 may select a subset of the message objects 1126 with the lowest n performance scores 1152 to assign to the restricted set 1162 and assign the remaining message objects 1126 to the allowed set 1160.
[0218] In some embodiments, the delivery controller 1114 may determine the assignment of the message object 1126 based on a comparison between the performance score 1152 and a threshold score. The threshold score may define the value of the performance score 1152 when assigning the corresponding message object 1126 to the permitted set 1160 or the restricted set 1162. If the performance score 1152 of the message object 1126 is determined to meet (e.g., be greater than or equal to) the threshold score, the delivery controller 1114 may assign the message object 1126 to the permitted set 1160. Conversely, if the performance score 1152 of the message object 1126 is determined to not meet (e.g., be less than) the threshold score, the delivery controller 1114 may assign the message object 1126 to the restricted set 1162. Based on the assignment to the permitted set 1160 or the restricted set 1162, the message selection system 1102 may determine whether to generate and transmit the message 1148 (e.g., in the manner described in Section D).
[0219] 11D, a sequence diagram of the system 1100 is depicted having operation of a message correlator 1116 within the message selection system 1102. The message correlator 1116 executing on the message selection system 1102 may identify one or more message objects 1126' of a second type (sometimes referred to herein as a rhetorical type). Each message object 1126' may include a set of executable instructions (e.g., a script) for evaluating whether to generate and send a corresponding message 1146' for presentation to a user 1132. In some embodiments, the message objects 1126' may correspond to the message objects 206 and their components described above in Sections B and C.
[0220] Each message object 1126' may include at least one message template 1142' and at least one selection criterion 1144', etc. The message template 1142' may include instructions for generating at least one corresponding message 1146' of a first type to include one or more user interface elements. For example, the message template 1142' may identify placeholders for inserting text or visual content into a graphical user interface element of a message to be presented via the remote computing device 1104. The selection criterion 1144' may define or include one or more endpoints to be achieved by the user 1132 when presenting the message 1146' corresponding to the message object 1126'. The selection criterion 1144' may also identify or define a time window within which the endpoints should be achieved. The endpoints of the selection criterion 1144' may include, for example, behavioral goals that are expected to be achieved by the user 1132. The selection criterion 1144' may also indicate which actions the user 1132 should take to satisfy one or more of the endpoints.
[0221] The message template 1142' of each message object 1126' of the second type may correspond to or be used to generate at least one message 1146'. For the message object 1126' of the second type, each message 1146' may have one or more non-interactive elements 1148'A-N (collectively referred to hereinafter as non-interactive elements 1148). Unlike the messages 1146 generated from the message object 1126 of the first type, the messages 1146' generated from the message object 1126' of the second type may lack one or more interactive elements 1150. Each non-interactive element 1148' may lack a configuration for responding to a detected interaction by the user 1132 when presented via the remote computing device 1104. Additionally, the non-interactive element 1148' may lack a configuration for recording an interaction by the user 1132. Because of such a configuration, the message 1146' may further lack the ability to send recorded interactions back to the message selection system 1102. For example, once the message 1146' is provided to the remote computing device 1104, the application 1128 may present the message 1146' to the user 1132. The application 1128 may parse the message 1146' and present (e.g., render or play) the message 1146' including the non-interactive elements 1148. While the message 1146' is being presented, a reporting agent 1130 of the application 1128 may monitor interactions with the message 1146'. Because the message 1146' lacks a configuration to report a response to the message selection system 1102, the reporting agent 1130 may prevent returning a response.
[0222] The message correlator 1116 may determine or identify at least one correspondence 1170 between one of the message objects 1126 of the first type in the permitted set 1160 and at least one message object 1126' of the second type. The correspondence 1170 may indicate that the message object 1126 and the message object 1126' have similar or matching endpoints as defined in the selection criteria 1144 and the selection criteria 1144', respectively. The identification of the correspondence 1170 may be based on the selection criteria 1144 of the message object 1126 of the first type and the selection criteria 1144' of the message object 1126' of the second type. In some embodiments, the message correlator 1116 may compare the selection criteria 1144 of each message object 1126 to the selection criteria 1144' of the message object 1126'. If the selection criteria 1144 are determined to not match, the message correlator 1116 may determine or identify a lack of correspondence 1170 between the message object 1126 and the message object 1126'. Additionally, the message correlator 1116 may identify another message object 1126 from the permitted set 1160 for comparison. In some embodiments, the message correlator 1116 may include the message object 1126' in the restricted list 1162.
[0223] Otherwise, when the selection criteria 1144 are determined to be met, the message correlator 1116 may determine or identify a correspondence 1170 between the message object 1126 and the message object 1126′. Upon identifying the correspondence 1170, the message correlator 1116 may select the message object 1126′ for inclusion in the permitted set 1160. In some embodiments, the message correlator 1116 may select a subset of the message objects 1126′ that are determined to have a correspondence 1170 with one of the message objects 1126 in the permitted set 1160. Upon selection, the message correlator 1116 may include the subset of the second type of message objects 1126′ in the permitted list 1160 along with the first type of message objects 1126. By inclusion in the permitted set 1160, the message object 1126′ may be permitted to be selected for generation and provision to the remote computing device 1104. In some embodiments, message objects 1126′ assigned to the permitted list 1160 may be made more likely (e.g., more highly weighted) to be selected for generation and provision for presentation via the remote computing device 1104.
[0224] Referring now to FIG. 12, a flow diagram of a method 1200 for selecting and sending different types of messages across a network environment using trust values is depicted. Method 1200 may be implemented using any of the components as detailed herein in conjunction with FIGS. 1-9D. In brief overview, a server may identify a first type of message object (1205). The server may determine a number of users that satisfy an endpoint (1210). The server may determine a performance score (1215). The server may rank the message objects by the performance score (1220). The server may allow or restrict the message object (1225). The server may identify a corresponding message object of a second type (1230).
[0225] More specifically, a server (e.g., message selection system 1102) may identify (1205) a first type of message object (e.g., message object 1126). Each message object may include a message template (e.g., message template 1142) and selection criteria (e.g., selection criteria 1144), etc. The message template may specify the generation of a corresponding message and user interface elements on the message. The selection criteria may define one or more endpoints to be achieved when presenting a message corresponding to the message object.
[0226] The server may determine the number of users (e.g., users 1132) that satisfy the endpoint (1210). To do so, the server may identify response data (e.g., response data 1124) that includes an entry (e.g., entry 1140) for each message object. From each entry, the server may identify actions performed by the user. Using the identifying information, the server may compare the actions to the endpoints specified by the message object. When an action matches an endpoint, the server may increment the number of users that satisfy the endpoint.
[0227] The server may determine (1215) a performance score (e.g., performance score 1154). Based on the number of users that satisfy the endpoint, the server may calculate a performance score for each message object. In some embodiments, the server may determine the performance score using a performance model (e.g., performance model 1116). The server may rank (1220) the message objects by their performance scores. The server may compare the performance scores of the message objects to a threshold.
[0228] The server may allow or restrict (1225) the message objects. Based on the ranking, the server may allow or restrict the message objects. For example, the server may select the top ranked message objects as allowed, while restricting the remaining message objects. Message objects identified as allowed may be allowed or made more likely to be used to generate and send the corresponding message (e.g., message 1146). Message objects identified as restricted may be prevented or made less likely to be used to generate and send the corresponding message.
[0229] The server identifies (1230) a corresponding message object of the second type (e.g., message object 1126'). The second type message object may have a message template (e.g., message template 1144') that specifies the absence of an interactive element (e.g., interactive element 1150). The server may identify the second type message object as corresponding to the first type message object if the endpoints of the two message objects match.
[0230] Although the present disclosure has been described with respect to specific embodiments, those skilled in the art will recognize that numerous variations are possible. For example, although specific examples of rules (including trigger conditions or resulting actions) and processes for generating suggested rules are described, other rules and processes can be implemented. Embodiments of the present disclosure can be implemented using a variety of computer systems and communication technologies, without being limited to the specific examples described herein.
[0231] The embodiments of the present disclosure may be realized using any combination of dedicated components or programmable processors or other programmable devices. The various processes described herein may be implemented on the same processor or different processors, in any combination. Where a component is described as being configured to perform a particular operation, such configuration may be achieved, for example, by designing an electronic circuit to perform the operation, by programming a programmable electronic circuit (such as a microprocessor) to perform the operation, or any combination thereof. Furthermore, while the embodiments described above refer to specific hardware and software components, one skilled in the art will appreciate that different combinations of hardware or software components may also be used, and that certain operations described as being implemented in hardware may also be implemented in software, or vice versa.
[0232] A computer program incorporating various features of the present disclosure may be encoded and stored on various computer readable storage media. Suitable media include magnetic disks or tapes, optical storage media such as compact disks (CDs) or DVDs (digital versatile disks), flash memory, and other non-transitory media. A computer readable medium encoding the program code may be packaged with a compatible electronic device, or the program code may be provided separately from the electronic device (e.g., downloaded over the Internet or provided as a separately packaged computer readable storage medium).
[0233] Although the specification contains many specific embodiment details, these should not be construed as limitations on the scope of any invention or what may be claimed, but rather as descriptions of features specific to particular embodiments of a particular invention. Certain features described herein in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented in multiple embodiments separately or in any suitable subcombination. Furthermore, although features may be described above as acting in a particular combination, and may even be initially claimed as such, one or more features from a claimed combination may in some cases be extracted from the combination, and the claimed combination may be directed to a subcombination or a variation of the subcombination.
[0234] Similarly, although the figures depict operations in a particular order, this should not be understood as requiring that such operations be performed in the particular order depicted, or sequentially, or that all of the depicted operations be performed, in order to achieve desirable results. In certain circumstances, multitasking or parallel processing may be advantageous. Furthermore, the separation of various system components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the program components and systems described above may generally be integrated into a single software product or packaged into multiple software products.
[0235] References to "or" may be interpreted inclusively such that terms described using "or" can refer to either a single, a plurality, or all of the described terms.
[0236] Thus, certain embodiments of the subject matter have been described. Other embodiments are within the scope of the following claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve desirable results. Moreover, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In certain embodiments, multitasking and parallel processing may be advantageous.
[0237] Although specific embodiments of the methods and systems have been described, it will be apparent to one skilled in the art that other embodiments incorporating the concepts of the present invention may be used. It should be understood that the above-described systems may provide any or more than one of their components, and that these components may be provided on either a stand-alone machine or, in some embodiments, on multiple machines in a distributed system. The above-described systems and methods may be implemented as methods, apparatus, or articles of manufacture using programming or engineering techniques to generate software, firmware, hardware, or any combination thereof. Furthermore, the above-described systems and methods may be provided as one or more computer-readable programs embodied on or within one or more articles of manufacture. As used herein, the term "article of manufacture" is intended to encompass code or logic embodied in and accessible from one or more computer-readable devices, firmware, programmable logic, memory devices (EEPROM, ROM, PROM, RAM, SRAM, etc.), hardware (e.g., integrated circuit chips, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), etc.), electronic devices, computer-readable non-volatile storage devices (CD-ROMs, floppy disks, hard disk drives, etc.). The article of manufacture may be accessible from a file server that provides access to the computer readable program via a network transmission line, a wireless transmission medium, a signal propagating through the air, radio waves, infrared signals, etc. The article of manufacture may also be a flash memory card or a magnetic tape. The article of manufacture includes hardware logic as well as software or programmable code embedded in a computer readable medium for execution by a processor. In general, the computer readable program may be implemented in any programming language such as LISP, PERL, C, C++, C#, PROLOG, or any bytecode language such as JAVA.The software programs may be stored as object code on or within one or more articles of manufacture.
[0238] Thus, although the disclosure has been described in terms of specific embodiments, it will be understood that the disclosure is intended to cover all modifications and equivalents that fall within the scope of the following claims.
Claims
1. 1. A method for transmitting a message over a network environment, comprising: identifying, by at least one server having one or more processors, a plurality of candidate message objects for selecting and generating corresponding candidate message objects, each of the plurality of candidate message objects being associated with selection criteria for presentation to a user to achieve one or more of a plurality of endpoints; obtaining, by the at least one server, from a user activity log associated with the remote computing device, one or more actions by the user recorded via the remote computing device; determining, by the at least one server, from the one or more recorded actions by the user, a user state used to identify one of the plurality of candidate message objects; determining, by the at least one server, a trust value for each message candidate of each of the plurality of message object candidates, the trust value being indicative of a predicted effectiveness of a corresponding message candidate for the user to achieve one or more of the plurality of endpoints based on a comparison of the user state to a selection criterion for the message object candidate; updating, by the at least one server, for each of the plurality of candidate message objects, the trust value based on at least one elapsed time from a sending time of a previous message corresponding to at least one of the plurality of candidate message objects to a current time; selecting, by the at least one server, a message object from the plurality of candidate message objects based on the trust value for the message; generating, by the at least one server, a message based on the message object selected from the plurality of candidate message objects; and transmitting, by the at least one server, the message for presentation to the remote computing device associated with the user; A method comprising:
2. The method of claim 1 further comprising: recording, by the at least one server, the activity log of one or more subsequent actions by the user in response to the submission of the message via the remote computing device; and updating, by the at least one server, a state of the user based on the one or more subsequent actions in response to the presentation of the message; A method comprising:
3. 2. The method of claim 1, further comprising the step of: establishing, by the at least one server, a message object model using a training data set including historical response data from the user to messages corresponding to one or more of the plurality of candidate message objects. Also equipped with The method, wherein determining the trust value further comprises applying the message object model to the user state and the selection criteria of the plurality of message object candidates to determine the trust value.
4. 2. The method of claim 1, wherein each message object of the plurality of candidate message objects includes a message template for generating a corresponding message, a model for determining the trust value, a time constraint for sending the corresponding message, and the selection criteria defining one or more of the plurality of endpoints to achieve within a defined time window.
5. 2. The method of claim 1, wherein determining the user state further comprises determining the user state corresponding to at least one of the plurality of endpoints from a plurality of user states of the user associated with the remote computing device.
6. 2. The method of claim 1, wherein updating the trust value comprises identifying and applying to the trust value a cool-down factor corresponding to the elapsed time between the last transmission time and the current time.
7. 2. The method of claim 1 , wherein the step of selecting a message object further comprises the step of selecting a message object from the plurality of candidate message objects based on a comparison of the trust value for the message object to a threshold, the threshold being defined by a delivery policy for sending a message.
8. 2. The method of claim 1, wherein obtaining the one or more actions comprises obtaining from the activity log of the user the one or more actions by the user in response to a previous message submission corresponding to one or more of the plurality of candidate message objects.
9. The method of claim 1 , wherein sending the message further comprises sending the message to the remote computing device according to a delivery policy for sending messages.
10. 2. The method of claim 1, further comprising removing, by the at least one server, from selection in response to the user state not satisfying the selection criteria for at least one message object candidate of the plurality of message object candidates.
11. 1. A system for transmitting messages across a network environment, comprising at least one server having one or more processors connected to a memory, the at least one server performing the following: identifying a plurality of candidate message objects for selecting and generating corresponding candidate message objects, each of the plurality of candidate message objects being associated with selection criteria for presentation to a user for achieving one or more of a plurality of endpoints; obtaining, from a user activity log associated with the remote computing device, one or more actions by the user recorded via the remote computing device; determining a user state from the one or more recorded actions by the user, the user state being used to identify one of the plurality of candidate message objects; determining, by the at least one server, for each message candidate of each of the plurality of message object candidates, a confidence value indicative of a predicted effectiveness of the corresponding message candidate for the user to achieve one or more of the plurality of endpoints based on a comparison of the user state to a selection criterion for the message object candidate; updating, for each of the plurality of candidate message objects, the confidence value based on at least one elapsed time from a sending time of a previous message corresponding to at least one of the plurality of candidate message objects to a current time; selecting a message object from the plurality of candidate message objects based on the trust value for the message; generating a message based on the message object selected from the plurality of candidate message objects; and sending the message for presentation to the remote computing device associated with the user; The system is configured as follows:
12. 12. The system of claim 11, wherein the at least one server further comprises: recording the activity log of one or more subsequent actions by the user in response to the presentation of the message via the remote computing device; and updating a state of the user based on the one or more subsequent actions in response to presentation of the message; The system is configured as follows:
13. 12. The system of claim 11, wherein the at least one server further comprises: establishing a message object model using a training data set including historical response data from the user to messages corresponding to one or more of the plurality of candidate message objects; and applying said message object model to said user state and said selection criteria of said plurality of candidate message objects to determine said confidence value; The system is configured as follows:
14. 12. The system of claim 11, wherein each message object of the plurality of candidate message objects includes a message template for generating a corresponding message, a model for determining the trust value, a time constraint for sending the corresponding message, and the selection criteria defining one or more of the plurality of endpoints to achieve within a defined time window.
15. 12. The system of claim 11, wherein the at least one server is further configured to determine the user state corresponding to at least one of the plurality of endpoints from a plurality of user states of the user associated with the remote computing device.
16. 12. The system of claim 11, wherein the at least one server is further configured to determine and apply a cool-down factor to the trust value corresponding to the elapsed time between the last transmission time and the current time.
17. 12. The system of claim 11, wherein the at least one server is further configured to select a message object from the plurality of candidate message objects based on a comparison of the trust value for the message object to a threshold, the threshold being defined by a delivery policy for sending messages.
18. 12. The system of claim 11, wherein the at least one server is further configured to retrieve from the activity log of the user, the one or more actions by the user in response to a previous message submission corresponding to one or more of the plurality of candidate message objects.
19. 12. The system of claim 11, wherein the at least one server is configured to send the message to the remote computing device according to a delivery policy for sending the message.
20. 12. The system of claim 11, wherein the at least one server is further configured to remove the at least one message object candidate from selection in response to the user state not satisfying the selection criteria for at least one message object candidate of the plurality of message object candidates.
Citation Information
Patent Citations
JPP6945909B
Optimized selection and delivery of content
US10242381B1
Methods and systems for optimizing messages to users of a social network
US20160269347A1
Adaptive nudge messages to motivate individuals to achieve certain wellness goals
US20170208021A1
Therapeutic system and remote patient monitoring device
US20180075219A1
