A system and method for preventing and warning about indulgence based on a causal discovery network
By generating personalized internet addiction warning messages and intervention suggestions through causal discovery networks and intelligent agents, the problems of high false alarm rates and lack of targeted intervention suggestions in existing systems are solved, achieving more efficient and accurate intervention for user internet addiction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- 湖南红普创新科技发展有限公司
- Filing Date
- 2026-02-10
- Publication Date
- 2026-05-12
AI Technical Summary
Existing internet addiction prevention systems cannot identify the reasons for user addiction at the causal logic level, resulting in a high false alarm rate and a lack of targeted intervention suggestions.
An anti-addiction early warning system based on causal discovery networks is adopted. The system generates an early warning model through a federated learning architecture, and uses causal discovery networks and intelligent agents to generate user attribution reports and intervention suggestions. It also combines behavioral, psychological and physiological data for personalized analysis.
It improves the targeting and reliability of internet addiction warning messages and intervention suggestions, reduces generation time, and increases efficiency.
Smart Images
Figure CN121684066B_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the fields of artificial intelligence technology and big data technology, and in particular relates to an anti-addiction early warning system and method based on causal discovery networks. Background Technology
[0002] With the popularization of the Internet and smart devices, internet addiction has become a global social issue. Internet addiction refers to the behavior of users who cannot control the duration and frequency of their Internet use, excessively rely on Internet services, and thus have a negative impact on their physical and mental health, social relationships, work, study and daily life.
[0003] To address this issue, existing internet addiction prevention systems have emerged. However, these systems all employ static rules for intervention, such as setting daily usage limits or forcibly disconnecting users from the internet at fixed times. Because this approach only superficially restricts current user behavior and fails to identify the causal reasons for user addiction, the intervention suggestions from these systems often lack specificity. This results in a high false alarm rate and a severe disconnect between intervention suggestions and actual needs, weakening the effectiveness of the internet addiction prevention systems. Therefore, generating timely warning messages and intervention suggestions for current users regarding internet addiction is a pressing technical problem that needs to be solved. Summary of the Invention
[0004] The purpose of this application is to provide an anti-addiction early warning system based on causal discovery networks, which aims to solve the technical problem of how to generate early warning messages and intervention suggestions for current users regarding internet addiction.
[0005] In a first aspect, embodiments of this application provide an anti-addiction early warning system based on a causal discovery network. The anti-addiction early warning system includes a central server and multiple client devices. The central server stores a global model with a federated learning architecture.
[0006] The central server is used to send the global model to multiple client devices;
[0007] Each client device receives the global model, trains the global model using local training samples, obtains the parameters of the local model, selects the difference between the parameters of the local model and the parameter set of the global model as the local model update value, and uploads the local model update value to the central server.
[0008] The central server receives the local model update data uploaded by each client device, uses an aggregation formula to aggregate the local model update data of each client device, generates an early warning model, and pushes the early warning model to the target client device among multiple client devices.
[0009] The target client device generates an attribution report for the current user based on a predefined method through the causal discovery network in the early warning model. Based on the current user's attribution report and the intelligent agent, it generates early warning messages and intervention suggestions for the current user's internet addiction.
[0010] In one possible implementation of the first aspect, the central server is used to receive the local model update volume uploaded by each client device, aggregate the local model update volume of each client device using an aggregation formula, generate an early warning model, and push the early warning model to a target client device among multiple client devices, including:
[0011] The central server receives the local model updates uploaded by each client device. It aggregates the local model updates from each client device using an aggregation formula to obtain the update parameters of the global model. The total loss function is used to obtain the total loss value of the global model. When the total loss value is less than a preset value, the global model is determined to meet the convergence condition. The update parameters of the global model are then imported into the global model architecture to generate an early warning model, which is then pushed to the target client devices among multiple client devices.
[0012] In one possible implementation of the first aspect, the target client device generates an attribution report for the current user through a causal discovery network in the early warning model, based on a predefined method. Based on the attribution report and the agent, it generates an early warning message and intervention suggestions for the current user regarding internet addiction, including:
[0013] The target client device receives the early warning model, deploys the early warning model, obtains the current user's behavioral data from the operation log, obtains the current user's psychological data from the psychological assessment table, and obtains the current user's physiological data from the data collected by the wearable smart device connected to the target client device.
[0014] The target client device uses a preset feature extraction model to extract features from the current user's behavioral data, psychological data, and physiological data, generating features for each. These features are then concatenated to obtain the current user's current fused features. These features are input into an early warning model, which uses a causal discovery network to generate a causal relationship for the user's internet addiction behavior. This causal relationship is then input into a counterfactual reasoning module within the early warning model, which generates intervention parameters for the user's internet addiction behavior. Finally, these intervention parameters are input into a generator within the early warning model, which generates an attribution report for the user.
[0015] The target client device obtains the reasons for the current user's internet addiction behavior and the current user's internet addiction level from the current user's attribution report. When the current user's internet addiction level is greater than the preset level, it generates an early warning message for the current user regarding internet addiction. The reasons for the current user's internet addiction behavior are input into the trained agent, which then generates intervention suggestions for the current user regarding internet addiction.
[0016] In one possible implementation of the first aspect, the current user's behavioral data includes one or more of the following:
[0017] The application's usage event sequence includes the application's startup timestamp, application's shutdown timestamp, application runtime, screen swiping speed, screen typing speed, URL categories in network access logs, access time periods in network access logs, and search keywords in network access logs. The application's usage event sequence is a collection of application operation events recorded by the target client device in chronological order.
[0018] In one possible implementation of the first aspect, the current user's mental data includes one or more of the following combinations:
[0019] The user's current negative emotional state rating, current stress level rating, and current loneliness rating;
[0020] Among them, the higher the current user's negative emotional state score, the current user's stress level score, and the current user's loneliness score, the higher the current user's internet addiction level.
[0021] Among them, the lower the current user's score for negative emotional state, stress level, and loneliness, the lower the current user's level of internet addiction.
[0022] In one possible implementation of the first aspect, the current user's physiological data includes one or more of the following combinations:
[0023] Current user's skin conductance value, current user's resting heart rate, current user's sleep duration;
[0024] Among them, the higher the current user's skin conductance value and the current user's resting heart rate, the higher the current user's level of internet addiction; the lower the current user's skin conductance value and the current user's resting heart rate, the lower the current user's level of internet addiction.
[0025] The shorter the current user's sleep duration, the higher the current user's internet addiction level; the longer the current user's sleep duration, the lower the current user's internet addiction level.
[0026] In one possible implementation of the first aspect, the training samples include multimodal data of a preset user and attribution reports of a preset user. The multimodal data of the preset user includes behavioral data, psychological data, and physiological data of the preset user. The attribution reports of the preset user include the reasons for the preset user's internet addiction behavior and the preset user's level of internet addiction.
[0027] The target client device is a client device that logs into an account belonging to a minor.
[0028] In one possible implementation of the first aspect, the aggregation formula is defined as follows:
[0029] );
[0030] in, This represents the update parameters of the global model; the update parameters of the global model are at the [number]th [time unit]. The set of parameters for the global model in the next round of training; Indicates the first The number of local model updates uploaded by each client;
[0031] Indicates the first The set of parameters for the global model during each round of training;
[0032] Indicates the learning rate; Indicates the serial number of the client device;
[0033] Indicates the first Data volume per client device This represents the total data volume, which is obtained by adding up the data volume of all client devices.
[0034] In one possible implementation of the first aspect, the total loss function is defined as follows:
[0035] ;
[0036] ;
[0037] This represents the total loss value of the global model. The larger the total loss value, the weaker the overall performance of the global model in terms of classification ability and causal reasoning ability; the smaller the total loss value, the stronger the overall performance of the global model in terms of classification ability and causal reasoning ability.
[0038] This represents the cross-entropy loss of the global model; the larger the cross-entropy loss of the global model, the weaker the classification ability of the global model; the smaller the cross-entropy loss of the global model, the stronger the classification ability of the global model.
[0039] The constraint loss of the global model represents the degree to which the learned causal graph structure deviates from the directed acyclic graph, and the weaker the causal reasoning ability. The smaller the constraint loss of the global model represents the degree to which the learned causal graph structure deviates from the directed acyclic graph, and the stronger the causal reasoning ability.
[0040] This represents a hyperparameter used to balance the constraint loss of the global model with the cross-entropy loss of the global model.
[0041] This represents the sum of all elements on the main diagonal of the matrix; The adjacency matrix represents the structure of a cause-effect graph. Represents the Hadamard product of the adjacency matrix; The matrix exponent is the Hadamard product of the adjacency matrix. This represents the dimension of the adjacency matrix.
[0042] Secondly, embodiments of this application provide an early warning method for the aforementioned anti-addiction early warning system, the early warning method comprising:
[0043] The target client device reads the preset push time and determines whether the current time is the push time;
[0044] If the current time is the push time, a warning message and intervention suggestions regarding the current user's internet addiction will be pushed to the monitoring terminal.
[0045] The beneficial effects of the embodiments of this application are as follows:
[0046] Firstly, since the target client device generates the current user's attribution report through the causal discovery network in the early warning model in a predefined manner, and generates early warning messages and intervention suggestions for the current user's internet addiction based on the current user's attribution report and the intelligent agent, the technical problem of how to generate early warning messages and intervention suggestions for the current user's internet addiction is solved. Since the intelligent agent automatically generates early warning messages and intervention suggestions for the current user's internet addiction, the generation time of early warning messages and intervention suggestions for the current user's internet addiction is reduced, which is conducive to improving the generation efficiency of early warning messages and intervention suggestions for the current user's internet addiction.
[0047] Secondly, based on the current user's attribution report and the agent, early warning messages and intervention suggestions for the current user's internet addiction are generated. Since the current user's attribution report contains the reasons for the current user's internet addiction behavior and the current user's level of internet addiction, the early warning messages and intervention suggestions for the current user's internet addiction can be derived based on the current user's attribution report. Therefore, it can improve the targeting and reliability of the early warning messages and intervention suggestions for the current user's internet addiction. Attached Figure Description
[0048] Figure 1 This is a schematic block diagram of the anti-addiction early warning system provided in the embodiments of this application;
[0049] Figure 2 This is an application scenario diagram of the anti-addiction early warning system provided in the embodiments of this application;
[0050] Figure 3 The application flowchart of the intelligent agent provided in the embodiments of this application;
[0051] Figure 4 This is a flowchart illustrating the implementation of the anti-addiction warning method provided in this application embodiment. Detailed Implementation
[0052] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.
[0053] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.
[0054] It should be understood that in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. The terms "comprising," "including," "having," and their variations all mean "including but not limited to," unless otherwise specifically emphasized.
[0055] Furthermore, the technical solutions of the various embodiments can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed in this application.
[0056] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.
[0057] Example 1
[0058] refer to Figure 1 , Figure 1 This is a schematic block diagram of the anti-addiction early warning system provided in the embodiments of this application, which is described in detail below:
[0059] The anti-addiction early warning system includes a central server and multiple client devices. The central server connects to the multiple client devices through a federated learning architecture, and stores a global model of the federated learning architecture.
[0060] The central server is used to send the global model to multiple client devices;
[0061] Each client device receives the global model, trains the global model using local training samples, obtains the parameters of the local model, selects the difference between the parameters of the local model and the parameter set of the global model as the local model update value, and uploads the local model update value to the central server.
[0062] The central server receives the local model update data uploaded by each client device, uses an aggregation formula to aggregate the local model update data of each client device, generates an early warning model, and pushes the early warning model to the target client device among multiple client devices.
[0063] The target client device generates an attribution report for the current user based on a predefined method through the causal discovery network in the early warning model. Based on the current user's attribution report and the intelligent agent, it generates early warning messages and intervention suggestions for the current user's internet addiction.
[0064] For ease of explanation, the following example is provided:
[0065] For example, the current user is user A;
[0066] User A's behavior: He secretly plays mobile games in his room after 11:30 p.m. every day, with each session lasting 2-3 hours, and sometimes even until 2 a.m. on weekends; he dozes off in class during the day, procrastinates on his homework, and communicates less with his parents.
[0067] User A's internet addiction level: High level;
[0068] The reasons for User A's internet addiction: Lack of supervision after family members fall asleep at night; the ranked matches and rewards in games make it hard to stop; staying up late leads to sleep deprivation, poor daytime energy, and low learning efficiency, leading to a reliance on late-night games to escape real-life pressures, creating a vicious cycle of playing late at night, feeling sleepy during the day, and wanting to play even more. Core conclusion: Unsupervised gaming behavior late at night is the direct cause of this teenager's internet addiction.
[0069] User A's warning message regarding internet addiction is as follows: Hello! The current time is 23:30, which is the nighttime rest period. We have detected that you are playing a game. We suggest you stop playing immediately and ensure you get 8-9 hours of sleep each night.
[0070] User A's late-night gaming time is the primary risk factor;
[0071] The intervention suggestion for User A regarding internet addiction is: reduce the time spent playing games late at night, and display the real-time game time during late nights in the youth mode interface so that User A can intuitively see the risk level of their behavior.
[0072] For ease of explanation, the following example is provided:
[0073] For example, the current user is user B;
[0074] User B's behavior: User B had an argument with a classmate and became irritable and easily angered. At 11:30 PM, User B logged into a team-based competitive mobile game and used the voice chat function to vent about workplace conflicts with game friends while participating in ranked matches. The game's built-in team-based star-free system made User B unwilling to leave the team even when feeling unwell. Combined with the real-time social interaction through the voice chat channel, the sense of time passing diminished, and User B ultimately played until 2:30 AM. During this time, the phone's screen time statistics function did not update the data in time due to the game running at a high priority, so User B could not visually see that the screen time limit had been exceeded.
[0075] User B's internet addiction level: High level;
[0076] The reason for User B's internet addiction: The instability of emotions caused by the user led User B to release emotions through social interaction through games, which affected normal rest and health.
[0077] User B received the following warning regarding internet addiction: Your current emotional state is mild anxiety, and playing games for more than 1.5 hours late at night will exacerbate this state. We recommend that you switch to mindfulness meditation audio to help you calm down more quickly.
[0078] The intervention recommendations for User B regarding internet addiction are as follows:
[0079] It is recommended to strengthen offline communication or psychological counseling. For example, when under pressure, don't use your phone alone to relieve stress. Instead, meet up with friends offline, chat about exam preparation experiences, and encourage each other. This will make it easier to soothe your emotions.
[0080] Client devices include, but are not limited to, mobile phones, tablets, in-vehicle devices, and laptops.
[0081] The number of client devices includes, but is not limited to, 4, 5, 6, 7, 8, 10, and 30. This application embodiment does not limit the number of client devices.
[0082] For ease of explanation, please refer to Figure 2 , Figure 2 This is an application scenario diagram of the anti-addiction early warning system provided in the embodiments of this application, which is described in detail below:
[0083] For example, there are multiple client devices, namely client device 1, client device 2, client device 3, and client device 4, and the central server connects client device 1, client device 2, client device 3, and client device 4 through a federated learning architecture.
[0084] The federated learning architecture eliminates the need to transfer massive amounts of data from client devices to a central server, reducing data transmission time and bandwidth consumption. When dealing with large-scale data, data transmission often becomes a bottleneck restricting model training speed. The federated learning architecture cleverly bypasses this problem through local training, significantly accelerating the training process of the global model.
[0085] The central server receives local model updates uploaded by each client device, aggregates these updates using an aggregation formula, generates an early warning model, and pushes the early warning model to a target client device among multiple client devices, including:
[0086] The central server receives the local model updates uploaded by each client device. It aggregates the local model updates from each client device using an aggregation formula to obtain the update parameters of the global model. The total loss function is used to obtain the total loss value of the global model. When the total loss value is less than a preset value, the global model is determined to meet the convergence condition. The update parameters of the global model are then imported into the global model architecture to generate an early warning model, which is then pushed to the target client devices among multiple client devices.
[0087] The training samples include multimodal data of preset users and attribution reports of preset users. The multimodal data of preset users includes behavioral data, psychological data, and physiological data of preset users. The attribution reports of preset users include the reasons for the preset users' internet addiction behavior and the preset users' internet addiction level.
[0088] The target client device is the client device that logs into the account of a minor.
[0089] The aggregation formula is defined as follows:
[0090] );
[0091] in, This represents the update parameters of the global model; the update parameters of the global model are at the [number]th [time unit]. The set of parameters for the global model in the next round of training; Indicates the first The number of local model updates uploaded by each client;
[0092] Indicates the first The set of parameters for the global model during each round of training;
[0093] Indicates the learning rate; Indicates the serial number of the client device;
[0094] Indicates the first Data volume per client device This represents the total data volume, which is obtained by adding up the data volume of all client devices.
[0095] The total loss function is defined as follows:
[0096] ;
[0097] ;
[0098] This represents the total loss value of the global model. The larger the total loss value, the weaker the overall performance of the global model in terms of classification ability and causal reasoning ability; the smaller the total loss value, the stronger the overall performance of the global model in terms of classification ability and causal reasoning ability.
[0099] This represents the cross-entropy loss of the global model; the larger the cross-entropy loss of the global model, the weaker the classification ability of the global model; the smaller the cross-entropy loss of the global model, the stronger the classification ability of the global model.
[0100] The constraint loss of the global model represents the degree to which the learned causal graph structure deviates from the directed acyclic graph, and the weaker the causal reasoning ability. The smaller the constraint loss of the global model represents the degree to which the learned causal graph structure deviates from the directed acyclic graph, and the stronger the causal reasoning ability.
[0101] This represents a hyperparameter used to balance the constraint loss of the global model with the cross-entropy loss of the global model.
[0102] This represents the sum of all elements on the main diagonal of the matrix; The adjacency matrix represents the structure of a cause-effect graph. Represents the Hadamard product of the adjacency matrix; The matrix exponent is the Hadamard product of the adjacency matrix. This represents the dimension of the adjacency matrix.
[0103] Optionally, after the central server receives the local model update data uploaded by each client device, aggregates the local model update data of each client device using an aggregation formula to generate an early warning model, and pushes the early warning model to the target client device among multiple client devices, before the target client device receives and deploys the early warning model, generates the current user's attribution report based on a predefined method, and generates early warning messages and intervention suggestions for the current user's internet addiction based on the current user's attribution report and the agent, the target client device is also used to integrate multiple risk states corresponding to the preset user's internet addiction level to form the state space of the reinforcement learning process, and integrate multiple preset suggestions corresponding to the reasons for the preset user's internet addiction behavior to form the action space of the reinforcement learning process. Using the reinforcement learning algorithm, the agent is trained based on the state space and action space to obtain the trained agent.
[0104] The beneficial effects of the embodiments of this application are as follows:
[0105] Firstly, since the target client device generates the current user's attribution report through the causal discovery network in the early warning model in a predefined manner, and generates early warning messages and intervention suggestions for the current user's internet addiction based on the current user's attribution report and the intelligent agent, the technical problem of how to generate early warning messages and intervention suggestions for the current user's internet addiction is solved. Since the intelligent agent automatically generates early warning messages and intervention suggestions for the current user's internet addiction, the generation time of early warning messages and intervention suggestions for the current user's internet addiction is reduced, which is conducive to improving the generation efficiency of early warning messages and intervention suggestions for the current user's internet addiction.
[0106] Secondly, based on the current user's attribution report and the agent, early warning messages and intervention suggestions for the current user's internet addiction are generated. Since the current user's attribution report contains the reasons for the current user's internet addiction behavior and the current user's level of internet addiction, the early warning messages and intervention suggestions for the current user's internet addiction can be derived based on the current user's attribution report. Therefore, it can improve the targeting and reliability of the early warning messages and intervention suggestions for the current user's internet addiction.
[0107] Example 2
[0108] Figure 3 The application flowchart of the intelligent agent provided in the embodiments of this application is described in detail below:
[0109] S301, the target client device receives the early warning model, deploys the early warning model, obtains the current user's behavioral data from the operation log, obtains the current user's psychological data from the psychological assessment table, and obtains the current user's physiological data from the data collected by the wearable smart device connected to the target client device;
[0110] S302, the target client device uses a preset feature extraction model to extract features from the current user's behavioral data, psychological data, and physiological data, respectively, generating features for the current user's behavioral data, psychological data, and physiological data. These features are then concatenated to obtain the current user's current fused features. These features are input into the early warning model, where the causal discovery network generates a causal relationship for the current user's internet addiction behavior. This causal relationship is then input into the counterfactual reasoning module of the early warning model, which generates intervention parameters for the current user's internet addiction behavior. Finally, these intervention parameters are input into the generator in the early warning model, which generates an attribution report for the current user.
[0111] S303, the target client device obtains the reason for the current user's internet addiction behavior and the current user's internet addiction level from the current user's attribution report. When the current user's internet addiction level is greater than the preset level, it generates an early warning message for the current user regarding internet addiction. It inputs the reason for the current user's internet addiction behavior into the trained agent and generates intervention suggestions for the current user regarding internet addiction through the trained agent.
[0112] The current user's behavior data includes one or more of the following combinations:
[0113] Among them, the application usage event sequence, the application startup timestamp, the application shutdown timestamp, the application runtime, the screen swiping speed, the screen typing speed, the URL categories in the network access logs, the access time periods in the network access logs, and the search keywords in the network access logs; the application usage event sequence is a collection of application operation events recorded by the target client device in chronological order;
[0114] For ease of explanation, the application uses event sequences, as exemplified below:
[0115] 1. Start the game, trigger matchmaking, and resume immediately after the match ends;
[0116] 2. Disable notifications on the target client device to block external interference.
[0117] 3. After playing continuously for 3 hours, switch to casual mode to play;
[0118] 4. Restart after 5 minutes to continue playing until the terminal is shut down. The current user's psychological data includes one or more of the following combinations:
[0119] The user's current negative emotional state rating, current stress level rating, and current loneliness rating;
[0120] Among them, the higher the current user's negative emotional state score, the current user's stress level score, and the current user's loneliness score, the higher the current user's internet addiction level.
[0121] Among them, the lower the current user's score for negative emotional state, stress level, and loneliness, the lower the current user's level of internet addiction.
[0122] The current user's physiological data includes one or more of the following combinations:
[0123] Current user's skin conductance value, current user's resting heart rate, current user's sleep duration;
[0124] Among them, the higher the current user's skin conductance value and the current user's resting heart rate, the higher the current user's level of internet addiction; the lower the current user's skin conductance value and the current user's resting heart rate, the lower the current user's level of internet addiction.
[0125] The shorter the current user's sleep duration, the higher the current user's internet addiction level; the longer the current user's sleep duration, the lower the current user's internet addiction level.
[0126] There is a complex and interconnected causal chain between current user behavioral data, psychological data, physiological data, and the occurrence of internet addiction. From the perspective of behavioral data, prolonged and frequent internet use, such as spending several hours daily immersed in online games or frequently browsing social media, is the most direct external manifestation of internet addiction and a direct factor leading to increased addiction. Continuous overuse strengthens the brain's dependence on online stimulation. Regarding psychological data, individual psychological traits such as low self-esteem, a tendency to escape real-world pressures, and a strong need for instant gratification drive users to actively seek solace in the online world, thus triggering internet addiction. Once addicted, the virtual sense of accomplishment and social satisfaction derived from the internet further solidify these negative psychological states, creating a vicious cycle. Based on current user physiological data, the physiological problems caused by long-term internet addiction, such as sleep deprivation, disrupted biological clock, and decreased vision, not only damage physical health but also have a reciprocal effect on psychology and behavior due to physical discomfort, making it more difficult for current users to break free from internet dependence. This is because poor physical condition may prompt them to use the internet to relieve pain or escape reality, ultimately making internet addiction behavior more and more serious.
[0127] The reasons for the current user's internet addiction behavior are input into the trained agent, which then generates intervention suggestions for the current user regarding internet addiction. Since these suggestions are tailored to the reasons behind the user's internet addiction behavior, they are personalized and tailored to the user's actual situation, thereby improving the targeting and effectiveness of the intervention suggestions.
[0128] Among them, the intervention suggestions for current users regarding internet addiction can help them clearly recognize their own problems, gradually change their bad online behaviors, effectively reduce the risk of internet addiction, and thus improve their quality of life and physical and mental health.
[0129] In this embodiment, the reasons for the current user's internet addiction behavior are input into the trained agent. The trained agent then generates intervention suggestions for the current user regarding internet addiction. By deeply analyzing the reasons for the current user's internet addiction behavior, the trained agent generates intervention suggestions that are tailored to the current user's actual situation, avoiding blind intervention and improving the accuracy and effectiveness of the intervention suggestions.
[0130] Example 3
[0131] refer to Figure 4 , Figure 4The following is a flowchart illustrating the implementation of the anti-addiction early warning method provided in this application embodiment, detailed below:
[0132] S401, The target client device reads the preset push time and determines whether the current time is the push time;
[0133] S402, if the current time is the push time, push the warning message and intervention suggestions on the current user's internet addiction to the monitoring terminal.
[0134] In this embodiment, if the current time is the push time, a warning message and intervention suggestions regarding the current user's internet addiction are pushed to the monitoring terminal. After receiving the warning message and intervention suggestions regarding the current user's internet addiction, the monitoring terminal can directly execute the intervention suggestions without secondary analysis, which greatly reduces the response delay and operational error of manual decision-making.
[0135] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and sub-samples of some embodiments may be included in or replace parts and sub-samples of other embodiments. Moreover, the terminology used herein is for descriptive purposes only and is not intended to limit the claims.
[0136] In this document, each embodiment focuses on describing the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.
[0137] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0138] The methods and products disclosed in the embodiments herein can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units may be merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some sub-samples may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to implement this embodiment according to actual needs. Furthermore, the functional units in the embodiments of this disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0139] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. Furthermore, any software tools or components not belonging to this company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.
Claims
1. An anti-addiction early warning system based on causal discovery networks, characterized in that, The anti-addiction early warning system includes a central server and multiple client devices. The central server stores a global model with a federated learning architecture. The central server is used to send the global model to multiple client devices; Each client device receives the global model, trains the global model using local training samples, obtains the parameters of the local model, selects the difference between the parameters of the local model and the parameter set of the global model as the local model update value, and uploads the local model update value to the central server. The central server receives the local model update data uploaded by each client device, uses an aggregation formula to aggregate the local model update data of each client device, generates an early warning model, and pushes the early warning model to the target client device among multiple client devices. The target client device receives the early warning model, deploys the early warning model, obtains the current user's behavioral data from the operation log, obtains the current user's psychological data from the psychological assessment table, and obtains the current user's physiological data from the data collected by the wearable smart device connected to the target client device. The target client device uses a preset feature extraction model to extract features from the current user's behavioral data, psychological data, and physiological data, generating features for each. These features are then concatenated to obtain the current user's current fused features. These features are input into an early warning model, which uses a causal discovery network to generate a causal relationship for the user's internet addiction behavior. This causal relationship is then input into a counterfactual reasoning module within the early warning model, which generates intervention parameters for the user's internet addiction behavior. Finally, these intervention parameters are input into a generator within the early warning model, which generates an attribution report for the user. The target client device obtains the reasons for the current user's internet addiction behavior and the current user's internet addiction level from the current user's attribution report. When the current user's internet addiction level is greater than the preset level, it generates an early warning message for the current user regarding internet addiction. The reasons for the current user's internet addiction behavior are input into the trained agent, which then generates intervention suggestions for the current user regarding internet addiction.
2. The anti-addiction early warning system as described in claim 1, characterized in that, The central server receives local model updates uploaded by each client device, aggregates these updates using an aggregation formula, generates an early warning model, and pushes the early warning model to the target client device among multiple client devices, including: The central server receives the local model updates uploaded by each client device. It aggregates the local model updates from each client device using an aggregation formula to obtain the update parameters of the global model. The total loss function is used to obtain the total loss value of the global model. When the total loss value is less than a preset value, the global model is determined to meet the convergence condition. The update parameters of the global model are then imported into the global model architecture to generate an early warning model, which is then pushed to the target client devices among multiple client devices.
3. The anti-addiction early warning system as described in claim 1, characterized in that, The current user's behavior data includes one or more of the following combinations: The application's usage event sequence includes the application's startup timestamp, application's shutdown timestamp, application runtime, screen swiping speed, screen typing speed, URL categories in network access logs, access time periods in network access logs, and search keywords in network access logs. The application's usage event sequence is a collection of application operation events recorded by the target client device in chronological order.
4. The anti-addiction early warning system as described in claim 1, characterized in that, The current user's psychological data includes one or more of the following combinations: The user's current negative emotional state rating, current stress level rating, and current loneliness rating; Among them, the higher the current user's negative emotional state score, the current user's stress level score, and the current user's loneliness score, the higher the current user's internet addiction level. Among them, the lower the current user's score for negative emotional state, stress level, and loneliness, the lower the current user's level of internet addiction.
5. The anti-addiction early warning system as described in claim 1, characterized in that, The current user's physiological data includes one or more of the following combinations: Current user's skin conductance value, current user's resting heart rate, current user's sleep duration; Among them, the higher the current user's skin conductance value and the current user's resting heart rate, the higher the current user's level of internet addiction; the lower the current user's skin conductance value and the current user's resting heart rate, the lower the current user's level of internet addiction. The shorter the current user's sleep duration, the higher the current user's internet addiction level; the longer the current user's sleep duration, the lower the current user's internet addiction level.
6. The anti-addiction early warning system as described in claim 1, characterized in that, The training samples include multimodal data of preset users and attribution reports of preset users. The multimodal data of preset users includes behavioral data, psychological data, and physiological data of preset users. The attribution report for preset users includes the reasons for the preset user's internet addiction behavior and the preset user's level of internet addiction. The target client device is a client device that logs into an account belonging to a minor.
7. The anti-addiction early warning system as described in claim 1, characterized in that, The aggregation formula is defined as follows: )； in, This represents the update parameters of the global model; the update parameters of the global model are at the [number]th [time unit]. The set of parameters for the global model in the next round of training; Indicates the first The number of local model updates uploaded by each client; Indicates the first The set of parameters for the global model during each round of training; Indicates the learning rate; Indicates the serial number of the client device; Indicates the first Data volume per client device This represents the total data volume, which is obtained by adding up the data volume of all client devices.
8. The anti-addiction early warning system as described in claim 2, characterized in that, The total loss function is defined as follows: ； ； This represents the total loss value of the global model. The larger the total loss value, the weaker the overall performance of the global model in terms of classification ability and causal reasoning ability; the smaller the total loss value, the stronger the overall performance of the global model in terms of classification ability and causal reasoning ability. This represents the cross-entropy loss of the global model; the larger the cross-entropy loss of the global model, the weaker the classification ability of the global model; the smaller the cross-entropy loss of the global model, the stronger the classification ability of the global model. The constraint loss of the global model represents the degree to which the learned causal graph structure deviates from the directed acyclic graph, and the weaker the causal reasoning ability. The smaller the constraint loss of the global model represents the degree to which the learned causal graph structure deviates from the directed acyclic graph, and the stronger the causal reasoning ability. This represents a hyperparameter used to balance the constraint loss of the global model with the cross-entropy loss of the global model. This represents the sum of all elements on the main diagonal of the matrix; The adjacency matrix represents the structure of a cause-effect graph. Represents the Hadamard product of the adjacency matrix; The matrix exponent is the Hadamard product of the adjacency matrix. This represents the dimension of the adjacency matrix.
9. A warning method based on the anti-addiction warning system according to claim 1, characterized in that, Early warning methods include: The target client device reads the preset push time and determines whether the current time is the push time; If the current time is the push time, a warning message and intervention suggestions regarding the current user's internet addiction will be pushed to the monitoring terminal.