Information processing device, information processing method, and program

Abstract

To provide an information processing device, an information processing method, and a program which are capable of taking action by understanding a relationship among a plurality of users.SOLUTION: An information processing device includes a score calculation unit configured to calculate a group trust score indicating a relationship among a plurality of users in a group in which at least two users are present from a plurality of different trust levels and change a user information handling level on the basis of the group trust score.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] In recent years, social robots that communicate with users and provide support have been developed. These social robots use user information to identify the users. User information includes, for example, face images, names, gender, age, information indicating the place of residence, and conversations with the social robot.

[0003] When communicating with a social robot, it is necessary to be careful about how the user's privacy information is handled. For this reason, for example, a method of re-editing text utterance data while preserving privacy and intent has been proposed, which is a method of generating similar output for processed understanding of fragments of the user's text data. This method refers to the procedure of re-editing or replacing with an alternative, and ensures privacy protection of data on highly sensitive topics (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] U.S. Pat. No. 1,113,8974 Summary of the Invention [Problem to be solved by the invention]

[0005] However, in the case of multiple users, the social robot may not understand the relationships between the users, and may respond in a general way without considering the context in which individuals in the group may trust each other.

[0006] The present invention has been made in consideration of the above problems, and has an object to provide an information processing device, an information processing method, and a program that are capable of understanding and responding to relationships between multiple users. [Means for solving the problem]

[0007] (1) In order to achieve the above-mentioned object, an information processing device according to one embodiment of the present invention is an information processing device that includes a score calculation unit that calculates a group trust score representing a relationship between multiple users in a group including at least two users from multiple different trust levels, and changes a handling level of user information based on the group trust score.

[0008] (2) In the information processing device according to the above (1) according to one aspect, the plurality of different trust levels may include three trust levels: social trust, family trust, and expert trust.

[0009] (3) In one embodiment of the information processing device of (1) or (2) above, the score calculation unit may change the weighting of each of the multiple different trust levels depending on the user's environment.

[0010] (4) In one aspect of the information processing device according to any one of (1) to (3), the score calculation unit may recognize the roles of the users, and recognize a host among the users based on the recognition result. If the host is recognized, the score calculation unit may calculate a reliability score of a relationship from the other users to the host for each pair of the host and a user other than the host, and calculate the group reliability score based on the reliability score calculated for each pair.

[0011] (5) In one aspect of the information processing device according to any one of (1) to (3), the score calculation unit may recognize the roles of the users, and recognize a host among the users based on the recognition result. If the host cannot be recognized, the score calculation unit may calculate a reliability score of a relationship between the pair of the users for each pair of the users, and calculate the group reliability score based on the calculated reliability score for each pair.

[0012] (6) In one embodiment of the information processing device of any one of (1) to (5) above, the handling level of the user information may be one of: permitting use of all of the user information, permitting use of part of the user information, or not permitting handling of all of the user information.

[0013] (7) In one embodiment of the information processing device of (4) or (5) above, one of the users may be a person selected by the information processing device, and the score calculation unit may calculate the reliability score of the selected person and calculate the group reliability score based on the calculated reliability score of the selected person.

[0014] (8) In the information processing device according to (4) or (5) of the present invention, an environmental sensor is provided for detecting environmental information of an environment in which the plurality of users are present, and the score calculation unit is The environment in which the multiple users are present may be estimated based on the environmental information detected by the environmental sensor, an environmental score for the environment in which the multiple users are present may be determined based on the estimated result, and the group reliability score may be calculated based on the determined environmental score.

[0015] (9) In order to achieve the above-mentioned object, an information processing method according to one embodiment of the present invention is an information processing method in which a score calculation unit calculates a group trust score representing a relationship between multiple users in a group including at least two users from multiple different trust levels, and changes a handling level of user information based on the group trust score.

[0016] (10) In order to achieve the above-mentioned object, a program according to one embodiment of the present invention is a program that causes a computer of an information processing device to calculate a group trust score representing a relationship between multiple users in a group containing at least two users from multiple different trust levels, and changes a handling level of user information based on the group trust score. Effect of the Invention

[0017] According to the above (1) to (10), the relationships between a plurality of users can be understood and handled accordingly. [Brief description of the drawings]

[0018] [Figure 1] FIG. 1 is a diagram for explaining an overview of an embodiment. [Diagram 2] FIG. 1 illustrates an example of a configuration of an information processing system according to an embodiment. [Diagram 3] FIG. 1 shows an example of a trust score between two individuals and a group trust score. [Figure 4] FIG. 13 is a diagram showing an example of a gathering setting, an example of a related gathering type, and an example of an environmental score. [Diagram 5] FIG. 2 is a diagram showing an example of a privacy profile stored in a storage unit. [Figure 6] 11 is a flowchart illustrating an example of a processing procedure for calculating a group reliability score according to the embodiment. [Figure 7] FIG. 11 is a diagram showing an example of a privacy policy stored in a storage unit. [Figure 8] FIG. 1 is a diagram illustrating an example of the configuration of an information processing system including a plurality of robots according to an embodiment. [Figure 9] FIG. 11 is a diagram showing an example of a privacy policy acquisition procedure according to the embodiment. [Figure 10] FIG. 13 shows an example of a user privacy policy. [Figure 11] 4 is a flowchart of a first process in the embodiment. [Figure 12] 4 is a flowchart of a first process in the embodiment. [Figure 13] 4 is a flowchart of a first process in the embodiment. [Figure 14] FIG. 13 is a diagram showing an example of a network and data set with a total of 109 connections when G=20 and J=8. [Figure 15] FIG. 13 is a diagram showing simple subgroups indicating trust relationships between users A, B, C, D, and E. [Figure 16] FIG. 16 shows an example of subgroup confidence scores based on FIG. 15 . [Figure 17] FIG. 17 is a diagram showing an example of a group reliability score based on FIG. 16. [Figure 18] FIG. 13 shows examples of initial reliability scores in the second evaluation. [Figure 19] FIG. 19 is a diagram showing an example of a group reliability score based on FIG. 18. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In the drawings used in the following description, the scale of each component is appropriately changed so that each component is of a recognizable size. In addition, in all the drawings for explaining the embodiments, the same reference numerals are used for the parts having the same functions, and the repeated explanation is omitted. In addition, "based on XX" in this application means "based on at least XX," and includes cases where it is based on other elements in addition to XX. Furthermore, "based on XX" is not limited to cases where XX is used directly, but also includes cases where it is based on XX that has been calculated or processed. "XX" is any element (for example, any information).

[0020] <Summary> First, an outline of this embodiment will be described. FIG. 1 is a diagram for explaining an outline of this embodiment. A robot 2 identifies the role of each of a plurality of participants (Hu1 to 4) within a predetermined range, and calculates a group trust score for each participant. The trust score is a score that indicates the relationship, for example, trust between two users, when at least two or more participants (users) are in a group. The group trust score is based on the trust score, and is a score that indicates the relationship between users in the group.

[0021] When there are multiple participants within a predetermined range, as in the case of symbol g1, the robot attempts to identify the multiple participants based on, for example, data detected by an environmental sensor or an audio signal picked up by a sound pickup unit.

[0022] Symbol g2 is an example of a recognition result. In this example, participants Hu1 to Hu3 were recognized (Yes), but participant Hu4 was not recognized (No).

[0023] Reference symbol g3 is an example of the result of recognizing the role of each participant. In this example, participant Hu1 is recognized as the host, participant Hu2 is recognized as the sibling of the host, participant Hu3 is recognized as the parent of the host, and participant Hu4 is recognized as a person unrelated to the host (or a person unknown to the host) (Stranger). Note that the recognition unit, for example, recognizes the host, and then recognizes the relationship between the other participants and the host.

[0024] Symbol g4 is an example of a reliability score. In this example, the score of participant Hu1 is high (High), the score of participant Hu2 is average (Avg), the score of participant Hu3 is above average (Above-Avg), and the score of participant Hu4 is low (Low). The score calculation unit calculates the reliability score between the robot 2 and the participants.

[0025] Thereafter, the score calculation unit performs an evaluation based on the reliability scores (symbol g5). The score calculation unit determines a group reliability score for the participants based on the evaluation results (symbol g6). The group reliability score affects interactions with participants, including the robot 2, after the determination, and the robot 2 (symbol g7).

[0026] Each participant has a person-to-person trust score formed from different criteria. The trust score helps to explain the trust relationship with someone else. The trust score between two people is determined by the individual's contact list and the trust score they have for each contact. In addition, if there are more than two people, it includes the calculation of a group trust score. The individual trust score (which does not refer to a specific individual) is similar to the recommendation trust score. In this embodiment, a group trust score is generated by processing all the person-to-person scores. The group trust score helps to determine the behavior of the robot 2, internal data processing, data sharing, and group interactions.

[0027] <Information Processing System> Next, a configuration example of the information processing system 1 of this embodiment will be described. Fig. 2 is a diagram showing a configuration example of the information processing system of this embodiment. As shown in FIG. 2, an information processing system 1 includes, for example, a robot 2 (information processing device) and an environmental sensor 3. The robot 2 includes, for example, an acquisition unit 201, an image processing unit 202, a sound processing unit 203, a motion processing unit 204, a recognition unit 205, a score calculation unit 206, a storage unit 207, a sensor 208, an actuator 209, and a policy processing unit 214. The robot 2 may include an image capturing unit 210 and a sound collecting unit 212. The robot 2 may also include an environmental sensor 3. The environment sensor 3 includes, for example, an image capturing unit 301, a sound collecting unit 302, and an output unit 303.

[0028] The image capturing unit 301 is, for example, an RBG-D image capturing device that can also acquire depth information D. Therefore, the image captured by the image capturing unit 301 includes image information and depth information.

[0029] The sound collection unit 302 is a microphone and collects a voice signal of a user. Note that the sound collection unit 302 may be a microphone array composed of a plurality of microphones. Note that in this embodiment, the collected voice signal is also one of the environmental information.

[0030] The output unit 303 outputs the image information captured by the image capture unit 301 and the audio signal collected by the audio collection unit 302 to the robot 2 as environmental information.

[0031] The environmental sensor 3 and the robot 2 are connected via a wireless or wired network NW. There may be two or more environmental sensors 3. The positions at which the environmental sensors 3 are installed may be known and may be stored in advance in the storage unit 207 of the robot 2. Alternatively, the environmental sensor 3 may output installation position information of the environmental sensor 3 to the robot 2 in association with environmental information.

[0032] The acquisition unit 201 receives the environmental information detected by the environmental sensor 3 .

[0033] The image capturing unit 210 is an image capturing device, such as a Charge Coupled Device (CCD) image capturing device or a Complementary Metal Oxide Semiconductor (CMOS) image capturing device, which is attached to the head or the like of the robot 2, for example.

[0034] The sound collection unit 212 is a microphone and collects a voice signal from a user. Note that the sound collection unit 212 may be a microphone array composed of a plurality of microphones.

[0035] The image processing unit 202 extracts, for example, an image of a person's face included in an image included in the environmental information detected by the environmental sensor 3 using a well-known method, and performs face recognition on the extracted face image using a well-known method. Alternatively, the image processing unit 202 extracts, for example, an image of a person's face included in an image captured by the imaging unit 210 using a well-known method, and performs face recognition on the extracted face image using a well-known method.

[0036] The voice processing unit 203 performs voice recognition processing (e.g., feature extraction processing, noise suppression processing, speech section estimation processing, sound source direction estimation processing, sound source estimation processing, etc.) on the voice signal picked up by the sound collection unit 302 or the sound collection unit 212 using well-known techniques.

[0037] The movement processing unit 204 controls the movement of the robot 2 in accordance with the program stored in the storage unit 207 and, for example, voice instructions from the user.

[0038] The identification unit 205 attempts to identify each of a plurality of people around the robot 2 using at least one of the results of the voice processing by the voice processing unit 203 and the results of the image processing by the image processing unit 202.

[0039] The score calculation unit 206 recognizes each person (user, participant) included in the captured image by using a known method. The score calculation unit 206 calculates a reliability score between people. The score calculation unit 206 calculates a group reliability score using the reliability scores between people.

[0040] The policy processing unit 214 performs processing related to the policy, for example, by using image-processed information and voice-recognition-processed information. Note that the privacy policy is a policy that determines whether the robot 2 may store user information, whether the information may be shared with other robots, etc. The policy processing unit 214 handles user information based on the privacy policy for each user stored in the storage unit 207.

[0041] The storage unit 207 stores programs, thresholds, mathematical expressions, etc. used for various controls of the robot 2. The storage unit 207 stores a privacy profile for each user. The storage unit 207 stores a privacy policy for each user.

[0042] The sensor 208 is a sensor, an encoder, etc., attached to each part of the robot.

[0043] The actuator 209 drives each part of the robot according to the control of the operation processing unit 204.

[0044] The robot 2 includes a power supply unit (not shown), and may also include an operation unit, an image display unit, a speaker, and the like.

[0045] <Reliability score> Next, the reliability score will be described. FIG. 3 shows an example of a trust score between two individuals and a group trust score. In this example, the trust score between two individuals H and G describes the level of trust that H has in G on different criteria. The trust score is calculated for a tuple of N H G =(t1,t2,…,t N ), where each value corresponds to a different measure of confidence and ranges from 0.0 to 1.0. Note that the confidence score is informed by many input sources (e.g., user-specified assignments, the results of previous interactions learned by Robot 2, estimates from similar associations, etc.).

[0046] In the example of FIG. 3, for example, three trust criteria are specified as reference examples. The three trust criteria used are, for example, social trust t s , family trust level t f , professional trust t p In Figure 3, the 3-tuple H G =(t s ,t f ,t p ) whose value ranges from 0.0 to 1.0. Then, the score calculation unit 206 ranks from no trust (0.0) to complete trust (1.0). In this way, the reliability score is calculated from a plurality of different trust levels (for example, social trust, family trust, and professional trust). Note that the trust levels are not limited to these three, but may be other trust levels, and the number is not limited to three, but may be two or more. Furthermore, the score calculation unit 206 may calculate the reliability score by changing the weighting for each of the plurality of different trust levels according to the user's environment.

[0047] In Figure 3, the first row shows the items, the second row shows the 3-tuple trust scores of H's coworkers, the third row shows the 3-tuple trust scores of H's siblings, the fourth row shows the 3-tuple trust scores of H's neighbors, and the fifth row shows the 3-tuple trust scores of H's strangers.

[0048] Here, the robot 2 may recommend an individual to the user. In this case, the individual may be unknown to the user, and the robot 2 may not have a reliability score associated with other individuals. The information processing system 1 may make a reliability recommendation to the user. For example, the information processing system 1 may try to recommend an academic teacher or colleague who is unknown to the user. Or, the information processing system 1 may try to recommend a bakery owner who specializes in making custom cakes that the user may need to celebrate a special event. In such cases, in this embodiment, the reliability is treated as an additive score used in the reliability calculation.

[0049] <Group Trust Score> In the following description, the following definitions are used. H is the "collection of all hosts at the time of calculation." |H| is the size or number of hosts in H. C is the "set of all contacts of the host." |C| is the size or number of contacts of C. Its value is C=H-1. Thus, when host y is taken from H and contact x is taken from C, y x corresponds to the trust score that y has for x. It is an n-tuple, with n distinct trust categories. Similarly, x y corresponds to the confidence score that x has for y, which is also an n-tuple with n distinct categories.

[0050] In addition, rec x and env score are n-tuples with the same distinct confidence intervals. RMS y corresponds to the "root mean square" of yx. It is a single number, i.e. (RMS y +1) is y x is the scalar to be multiplied by. (R.M.S. y +1)*y x , x y ,rec x Since all of the n-tuples are n-tuples, their sum also generates another n-tuple. Next, the score calculation unit 206 calculates score and the generated sum to generate another n-tuple.

[0051] This process is repeated for all possible combinations of hosts and contacts, and the results are then all summed. Note that the final result will be another n-tuple. Finally, we multiply this n-tuple by the scalars (1 / |H|) and (1 / |C|). This final value is the group trust score in the absence of the specific host.

[0052] So, when looking at the calculations when a dedicated host is present, the process is simplified. Many of the same principles apply, but the summation is only done over the results for all contacts of the dedicated host. Another exception is that the final summation now has a scalar multiplication by (1 / |C|) instead of what was done before.

[0053] <Environment detection> The Robot 2 may also use additional context about its environment to determine more information about the types of interactions that may occur between individuals while in the space with the Robot 2. The environment detection may be performed on the fly or may be fixed during the initial setup of the Robot 2, for example when set up at home, in an office, at school, in a public space, etc.

[0054] The environmental sensing score describes Robot 2's perception or belief about the type of gathering along three different axes: social gathering, family gathering, and professional gathering. For example, Robot 2 installed in a corporate office will understand that gatherings and interactions between users and participants in that location are also professional in nature. In this example, Robot 2 can assume that relationships between individuals in that location are most commonly professional in nature.

[0055] FIG. 4 is a diagram showing an example of a gathering setting, a related gathering type, and an example of an environment score. In FIG. 4, the first column is the type of group, the second column is the type of group, and the third column is the environment score. The first row is the item, the second row is the group at work, the third row is the group at school, the fourth row is the group at home, and the fifth row is the group at a public place. The group at work has a group type of professional and an environment score of (0.0, 0.0, 1.0). The group at school has a group type of professional or social and an environment score of (0.5, 0.0, 0.7). The group at home has a group type of family or social and an environment score of (0.7, 0.7, 0.0). The group at a public place has a group type of social and an environment score of (1.0, 0.0, 0.0). In this way, the environment score env score is a three-tuple (social trust t s , family trust level t f , professional trust t p ) It should be noted that, computationally, the environment score is weighted, for example, during the calculation of the group credibility score.

[0056] <Privacy Profile> Here, we explain the privacy profile. The privacy profile is information about the user. Fig. 5 is a diagram showing an example of a privacy profile stored in the storage unit. As shown in Fig. 5, the storage unit 207 stores a privacy profile in association with the user's identification information (user ID). The privacy profile is, for example, information on the location where the robot 2 is used (e.g., latitude and longitude, address, building name, company name, etc.), the user's age, occupation, sex, hobbies, etc. The privacy profile may also include the user's speech content (audio signal, audio feature amount, text data of the speech content, etc.), the user's facial photo information, etc. 5 is an example, and other information may be included, and information may be added, deleted, or updated. For example, the policy processing unit 214 processes such information.

[0057] <Processing example> In this embodiment, a general physical setup is assumed to be placed in a common space where the robot 2 can identify human individuals. The robot 2 may be stationary or mobile to perform this task (recognizing and identifying individuals). FIG. 6 is a flowchart showing an example of a processing procedure for determining a group reliability score according to this embodiment. In the following processing, it is assumed that there are two or more people around the robot 2, and that, for example, a conversation (activity) is taking place.

[0058] (Step S1) The environmental sensor 3 detects, for example, environmental information around the robot 2. The acquisition unit 201 acquires the environmental information from the environmental sensor 3. The score calculation unit 206 estimates the environment in which multiple users and the robot 2 exist based on, for example, images and position information included in the acquired environmental information, and calculates an environmental score env (env) based on the estimation result, for example, as shown in FIG. score The environment score may be determined in advance for each environment and stored in the storage unit 207.

[0059] (Step S2) Using the environmental information, the identification unit 205 attempts to identify individuals participating in an activity at the start of or during the human activity. If the identification unit 205 is able to identify the individual (step S2; YES), the identification unit 205 proceeds to processing in step S3. If the identification unit 205 is unable to identify the individual (step S2; NO), the identification unit 205 proceeds to processing in step S5. Note that, in the case of an individual not registered in the memory unit 207, the identification unit 205 determines that the individual is a stranger (unknown person) to the information processing system 1.

[0060] (Step S3) The identification unit 205 determines that the identification has been successful if the identified individual has already been registered in the storage unit 207. After the process, the identification unit 205 proceeds to the process of step S4.

[0061] (Step S4) The score calculation unit 206 searches the privacy profile of each identified individual from the storage unit 207. After the process, the score calculation unit 206 proceeds to the process of step S6.

[0062] (Step S5) The score calculation unit 206 regards those who could not be recognized as unknown users. After the process, the score calculation unit 206 proceeds to the process of step S6.

[0063] (Step S6) The score calculation unit 206 determines a particular individual as a host in the group based on the privacy profile, which may be designated by whether or not the individual is an active speaker or activity leader, among other possibilities.

[0064] (Step S6) If the score calculation unit 206 can determine the host (step S5; YES), it determines a pairwise trust score from other participants in the group to the host. That is, the score calculation unit 206 calculates a pairwise trust score for other participants with respect to the host. As described above, the trust score embodies the trust level within the group for each of the three trust types (social trust, family trust, and professional trust). After the process, the score calculation unit 206 proceeds to the process of step S8.

[0065] (Step S7) If the host cannot be determined (Step S5; NO), the score calculation unit 206 determines pairwise scoring for each of the participants in the group. That is, the score calculation unit 206 determines the pairwise reliability scoring between each pair of individuals in all permutations. After the process, the score calculation unit 206 proceeds to the process of step S8.

[0066] (Step S8) The score calculation unit 206 generates a group reliability score from all the pairwise scores.

[0067] This result is then used to guide Robot 2's behavior, taking into account group dynamics and trust within the group. The above-described process steps and process contents are merely examples, and are not limited to these. For example, some of the processes may be performed in parallel.

[0068] <Calculation of group trust score> Next, an example of a method for calculating a group reliability score will be described. When the host is determined, the score calculation unit 206 calculates the group trust score GroupTrust by using the following formula (1): H Calculate.

[0069]

number

[0070] In equation (1), H represents the set of all temporary hosts. C represents the set of all other individuals in the group. H x is the trust score that H has for contact x. RMS H is H x This is the root mean square calculation of rec x is the recommendation trust score of individual x. score is the environmental score. x H is the reliability score of x for H. Note that the recommendation reliability score is a three-tuple (social reliability t s , family trust level t f , professional trust t p) In other words, the recommendation reliability score is a reliability score for an individual recommended by the robot 2. The score calculation unit 206 calculates or determines the recommendation reliability score, which is a lookup value determined from the user profile of the individual.

[0071] Such a host-specified group trust score is calculated for each of the various criteria, representing the trust associated with each criterion for this host's group. A higher score indicates a higher degree of trust for this host when it comes to participating in a particular criterion. A lower score indicates a lower degree of trust for this host.

[0072] If the score calculation unit 206 is unable to determine the host, it calculates the group trust score GroupTrust using the following formula (2).

[0073]

number

[0074] In formula (2), y x represents the trust score that y has for contact x. RMS y is y x is the root mean square of x y is the confidence score of x for y.

[0075] Such a general group trust score creates a value for each of the various criteria, describing the trust associated with each criterion among all members of the group. Higher scores indicate more trust in engaging with that particular criterion. Lower scores indicate less trust in engaging with that particular criterion.

[0076] <Privacy Policy> Fig. 7 is a diagram showing an example of a privacy policy stored in the storage unit. As shown in Fig. 7, the storage unit 207 stores privacy policy information in association with user identification information (user ID). The privacy policy information includes information such as whether sharing is permitted or not permitted, whether retention is permitted or not permitted, and whether learning is permitted or not permitted. Note that the user ID may be associated with image information (face image, image feature amount, etc.) and voice information (voice feature amount, etc.).

[0077] <Information processing system with multiple robots> Fig. 8 is a diagram showing an example of the configuration of an information processing system including a plurality of robots according to this embodiment. In the example of Fig. 8, the number of robots is two, but the number of robots may be three or more. The information processing system 1A includes, for example, a first robot 2-1, a second robot 2-2, and an environmental sensor 3.

[0078] The configurations of the first robot 2-1 and the second robot 2-2 are similar to that of the robot 2 of the information processing system 1. For this reason, the first robot 2-1 includes, for example, an acquisition unit 201-1, an image processing unit 202-1, a sound processing unit 203-1, a motion processing unit 204-1, a recognition unit 205-1, a score calculation unit 206-1, a storage unit 207-1, a sensor 208-1, an actuator 209-1, a policy processing unit 214-1, an image capture unit 210-1, and a sound collection unit 212-1. The second robot 2-2 includes, for example, an acquisition unit 201-2, an image processing unit 202-2, a sound processing unit 203-2, a motion processing unit 204-2, a recognition unit 205-2, a score calculation unit 206-2, a storage unit 207-2, a sensor 208-2, an actuator 209-2, a policy processing unit 214-2, an image capturing unit 210-2, and a sound collection unit 212-2. Note that the configurations of the first robot 2-1 and the second robot 2-2 may be the same or different.

[0079] The first robot 2-1 is used by a first user and is installed, for example, in the home of the first user. The second robot 2-2 is used by a second user and is installed, for example, in the second user's home.

[0080] <Policy processing> Next, the process performed by the policy processing unit will be described. Fig. 9 is a diagram showing an example of a privacy policy acquisition procedure according to this embodiment. The following process is performed, for example, when the second robot 2-2 starts an interaction with a user.

[0081] (Step S101) The policy processing unit identifies people in a location based on, for example, an image captured by at least one of the image capture units 301, 210-1, or 210-2, or a sound captured by at least one of the sound capture units 302, 212-1, or 212-2. When there are multiple people, the policy processing unit 214-2 recognizes each person.

[0082] (Step S102) If the person has already been recognized and stored in the storage unit 207-2, the policy processing unit 214-2 identifies the person as a recognized person. Note that the person's identification information may be stored in a server or the like via the network NW.

[0083] (Step S103) The policy processing unit 214-2 permits collection and storage of a privacy profile, which is user information of the recognized individual, and associates the collected personal information with identification information and stores it in the storage unit 307. The user information is, for example, name, place of use, age, sex, occupation, place of residence, hobbies, and speech content. After completing step S103, the policy processing unit 214-2 proceeds to the process of step S106.

[0084] (Step S104) If the person has not already been recognized and stored in the storage unit 207-2, or if the individual cannot be identified, the policy processing unit 214-2 determines that the person is unknown to the robot.

[0085] (Step S105) Policy processing unit 214-2 creates the above-mentioned dummy privacy profile. After step S105 ends, policy processing unit 306 proceeds to the process of step S106.

[0086] (Step S106) The policy processing unit 214-2 loads the privacy profile of step S103 or step S105.

[0087] (Step S107) The policy processing unit 214-2 waits for data to be processed.

[0088] The above process allows the robot to understand the appropriate level of processing it can perform and what it should (or should not) do with personal user data.

[0089] <Personal Privacy Profile> First, the handling of privacy profiles and policies in this embodiment will be described. FIG. 10 is a diagram showing an example of a user public privacy policy.

[0090] The box g200 (third information area) contains, for example, public accessible user information g201 and a device address g202. The information in this area is personal information that may be appropriate to share with other robots. This includes, for example, information about the user (general user information), such as the user's name and user ID (identification information) within the robot system. It also includes, for example, device addresses that other robots can potentially use to send data.

[0091] The box g210 (first information area) includes, for example, a general policy g211, sensitive topics g212, and other robot permissions g213. The confidential topic g212 includes M (M is an integer equal to or greater than 2) topics 1 (g214-1) to M (g214-M). Each topic includes an estimated sensitivity score (Sensitivity Est. Score).

[0092] Note that Permissions for other robots (g213) is a generalized permission block that can be considered as the default policy for how other robots may use this person's data. This block contains a set of permission policies that determine the level of data processing that other robots may perform on this user.

[0093] Other robot permissions (g213) include, for example: a) Can I process my data (g215)? Can other robots (second robots) process this user data (yes or no)? b) Store process (Can Store) (g126): Whether or not another robot (second robot) can store this user information in the memory of the second robot (yes or no). c) Can Lean from my data (anon) (g217); Whether other devices (second robots) can learn from this user information (unrestricted, anonymously, or not) d) Can Share with others (g218): Whether other robots (second robots) can share this user information with the first robot of the user (yes or no). e) Can Send results to my device (g219): whether other robots (second robots) can send this user information back to the first robot of the user (send both results and raw data, send only results after processing, send only raw data before processing, or send nothing). The above-mentioned policies, processes, and judgments are merely examples and are not limited to these.

[0094] The purpose of Sensitive Topics (g212) is to inform the user of the sensitivity of a particular topic when it is discussed by another robot (the second robot). This policy could be, for example, whether it is an appropriate topic to discuss medical information such as allergies.

[0095] Additionally, the score on each topic provides granularity in determining permission to discuss the topic.

[0096] The box marked with the symbol g250 (second information area) contains a User Specific Policy g215, Contacts g252, M contacts 1 (g253-1) to M (g253-M), a Trust Score g254, Sensitive Topics g255, and Other Device Permissions g256. The confidential topic g255 includes M (M is an integer equal to or greater than 2) topics 1 (g257-1) to M (g257-M). Each topic includes an estimated sensitivity score (Sensitivity Est. Score). An example of the configuration and processing of the permission g256 for other robots will be described later.

[0097] Thus, the present embodiment may include personalized user-specific policies for customizing preferences when interacting with other particular individuals and their devices.

[0098] Also, there are permission blocks designated based on the contacts specified by the user, such as g250. Below these permission blocks, there may be permission blocks when the user interacts with a robot belonging to one of the contacts specified by the user. The reliability score g254 is a group reliability score, and the handling level of the user information is changed based on this group reliability score. Note that the handling level (processing level) of the user information is, for example, partial (PARTIAL) and full (FULL), as described later with reference to Figs. 11 to 13, and indicates that all use of the user information is permitted, that some use of the user information is permitted, and that handling of all the user information is not permitted. Note that the handling level is not limited to this.

[0099] In addition to the permission block and the sensitive topic block, users can provide some additional information about their contact, such as a trust score and other metadata, as shown in g250. Note that the structure, processing, and decision-making of other robots' permissions (g256) are similar to other robots' permissions (g213).

[0100] The structure of the privacy profile in this embodiment allows flexibility and extensibility, with entries being added, modified, and removed over time, depending on individual preferences and ongoing interaction with the robot.

[0101] <Default Privacy Profile and Dummy Privacy Profile> Next, we will explain the default privacy profile and the dummy privacy profile. Initially, when a user sets up the robot, a "default privacy profile" is created that contains, for example, the user's name, user ID, and device ID in the user information block. In the general permissions block, each individual policy is set to "NO" (not allowed) until the robot learns the user's preferences for each policy. Similarly, sensitive topic blocks are left empty until the robot learns about the user's specific topics and how to evaluate the sensitivity of such topics, and the robot leaves certain allowed blocks empty until the user designates a trusted contact.

[0102] The concept of a default privacy profile is further abstracted, and there is also the concept of a "dummy privacy profile" that can be used to represent a stranger to the robot's system. In this embodiment, we also use the concept of this "dummy privacy profile." The purpose of this "Dummy Privacy Profile" is to represent how the robot should handle and process user data of an individual who is unknown to the robot and the robot system. The content of the "Dummy Privacy Profile" is the same as the Default Privacy Profile, with the exception that there is no device ID associated with this individual (the username field is also left blank). Thus, the only content of the "Dummy Privacy Profile" is a uniquely generated user ID, and a general permission block with all individual policies set to "NO".

[0103] <Data processing method, data processing procedure> Next, an example of a data processing method and an example of a data processing procedure will be described. The workflow for each user, interaction data describes, for example, the "appropriate processing" to be performed by the second robot 2-2 and the information to be sent to the user's own first robot 2-1. And there are, for example, up to five checks that may be performed on the user information, which correspond to the device permissions of the user's privacy profile. In the following description, it is assumed that the robot that makes the decisions and processing is the second robot 2-2, and the robot that exchanges information is the first robot 2-1.

[0104] (First process) 11 is a flowchart of a first process in this embodiment. The first process is performed when the data processing check fails.

[0105] (Step S211) The policy processing unit 214-2 acquires processing data. The processing data is, for example, information based on a captured image of the user or the contents of a speech from a picked-up voice.

[0106] (Step S212) The policy processing unit 214-2 checks whether the second robot 2-2 can process the processing data (is permitted to process). If the second robot 2-2 can process the processing data (step S212; YES), the policy processing unit 214-2 proceeds to processing of step S221 (FIG. 12). If the second robot 2-2 cannot process the processing data (step S212; NO), the policy processing unit 214-2 proceeds to processing of step S213.

[0107] (Step S213) The policy processing unit 214-2 checks whether the processing data should be sent to the user's own first robot 2-1 for processing. If the processing data should be sent to the user's own first robot 2-1 for processing (Step S213; YES), the policy processing unit 214-2 proceeds to processing in step S214. If the processing data should not be sent to the user's own first robot 2-1 for processing (Step S213; NO), the policy processing unit 214-2 proceeds to processing in step S216.

[0108] (Step S214) Policy processing unit 214-2 encrypts the data to be processed. After the processing, policy processing unit 214-2 proceeds to the process of step S215.

[0109] (Step S215) The policy processing unit 214-2 transmits the encrypted processing data to the first robot 2-1 of the user using, for example, the device ID of the user. After the processing, the policy processing unit 214-2 ends the processing.

[0110] (Step S216) Policy processing unit 214-2 deletes the processed data without storing it. After the processing, policy processing unit 214-2 ends the processing.

[0111] (Second Processing) 12 is a flowchart of the second process in this embodiment. The second process is performed when the data processing check is successful.

[0112] (Step S212) If the second robot 2-2 can process the processing data (step S212; YES), the policy processing unit 214-2 proceeds to the processing of step S221.

[0113] (Step S221) Policy processing unit 214-2 determines the processing level. The processing level is, for example, partial (PARTIAL) and full (FULL). If the processing level is partial (PARTIAL), policy processing unit 214-2 proceeds to processing of steps S222 and S223. If the processing level is full (FULL), policy processing unit 214-2 proceeds to processing of step S225.

[0114] (Step S222) The policy processing unit 214-2 checks the metadata of the data to be processed by comparing the topics or keywords with the user's confidential topic list stored in the storage unit 207-2. Based on the result of the check, the policy processing unit 214-2 decides whether or not to permit data processing. If the policy processing unit 214-2 permits data processing (step S222; Permit), the process proceeds to step S224. If the policy processing unit 214-2 does not permit data processing (step S222; Not Permit), the process proceeds to step S224.

[0115] (Step S223) The policy processing unit 214-2 checks against the identified user and refers to the user's unique profile stored in the memory unit 207-2 regarding the processing of the user at the first robot 2-1. Note that this processing determines whether or not the processing of data depends on the trust relationship between the owner of the second robot 2-2 and the user. The result is whether the data processing is permitted or not. If permitted (step S223; permitted), the policy processing unit 214-2 proceeds to processing of step S224. If not permitted (step S224; not permitted), the policy processing unit 214-2 proceeds to processing of step S24.

[0116] (Step S224) The policy processing unit 214-2 checks concurrency based on the trust between the topic and other individuals. If the check passes (step S224; PASS), the policy processing unit 214-2 proceeds to processing in step S225. If the check fails (step S224; FAIL), the policy processing unit 214-2 proceeds to processing in step S213 (FIG. 11).

[0117] (Step S225) The policy processing unit 214-2 causes the second robot 2-2 to proceed to each process (steps S213, S231 (FIG. 13), S232 (FIG. 13), S233 (FIG. 13)) so as to process the process data.

[0118] (Third Processing) 13 is a flowchart of the third process in this embodiment. The third process is a process performed on the second robot 2-2.

[0119] (Step S225) The policy processing unit 214-2 advances to each process (steps S216, S231, S232, S233) so that the second robot 2-2 processes the processing data. The policy processing unit 214-2 performs the process of step S213 to determine whether the results can be sent to the first robot 2-1, which is the host device.

[0120] (Step S231) The policy processing unit 214-2 checks whether the processing result can be shared with the first robot 2-1 in order to notify the second robot 2-2 of its operation. If the processing result can be shared (step S231; YES), the policy processing unit 214-2 proceeds to processing of step S234. If the processing result cannot be shared (step S231; NO), the policy processing unit 214-2 proceeds to processing of step S216.

[0121] (Step S232) The policy processing unit 214-2 checks whether the processing result can be stored in the memory unit 207-2 of the second robot 2-2. If the processing result can be stored in the memory unit 207-2 of the second robot 2-2 (step S232; YES), the policy processing unit 214-2 proceeds to processing in step S235. If the processing result cannot be stored in the memory unit 207-2 of the second robot 2-2 (step S232; NO), the policy processing unit 214-2 proceeds to processing in step S216.

[0122] (Step S233) The policy processing unit 214-2 checks whether the second robot 2-2 can learn from the data and results anonymously. If the second robot 2-2 can learn from the data and results anonymously (step S233; YES), the policy processing unit 214-2 proceeds to processing of step S236. If the second robot 2-2 cannot learn from the data and results anonymously (step S233; NO), the policy processing unit 214-2 proceeds to processing of step S216.

[0123] (Step S216) If the check results in steps S231, S232, and S233 are NO, policy processing unit 214-2 deletes the results and data. After processing, policy processing unit 214-2 ends the process.

[0124] (Step S234) The policy processing unit 214-2 transmits the data and the result to the behavior processing unit 204-1 (Host's Device Behavior) of the first robot 2. After the processing, the policy processing unit 306 ends the processing.

[0125] (Step S235) The policy processing unit 306 transmits the data and the result to the first robot 2-1 so that they are stored in the storage unit 207-1 (Host's Device Storage) of the first robot 2-1.

[0126] (Step S236) The policy processing unit 306 transmits the data and the result to the operation processing unit 204-1 (Host's Device Logic) of the first robot 2-1. After the processing, the policy processing unit 214-2 ends the processing.

[0127] 11 to 13 are merely examples, and the present invention is not limited to these. For example, some of the processes may be performed simultaneously.

[0128] In the above example, the user's robot is the first robot 2-1 and the second robot 2-2 performs processing related to the policy, but the user's robot may be the second robot 2-2 and the first robot 2-1 may perform processing related to the policy.

[0129] <Evaluation> (First evaluation) In the first evaluation, we used semi-random datasets to test the rigor of our computations. The datasets were randomly generated with the following constraints: G is the total number of unique participants. J = the maximum number of contacts that a member of G can have. The final number is chosen randomly. The selection of contacts is done randomly.

[0130] In addition, reliability scores were randomly determined according to the following criteria: t1 is a random number between 0.5 and 0.9 t2 is a random number between 0.1 and t1 t3=1-t1-t2 These three values ​​are the 3-tuple of confidence scores (t s ,t f ,t p ) were randomly placed in the . The randomization range of values ​​in allowed for the situation where one of the reliability scores dominates the other two, potentially resulting in a secondary reliability score below average.

[0131] Figure 14 shows an example of a network data set with all 109 connections when G = 20 and J = 8. This directed graph shows whether there is a trust score from one individual to another. That is, the relationships connected by arrows indicate the existence of a trust relationship.

[0132] Figure 15 shows a simplified subgroup diagram showing the trust relationships between users A, B, C, D, and E. Figure 15 is also an example of a subgroup selected from the larger network in Figure 14. In the following scenario, individuals A, B, C, D, and E are interacting with Robot 2, which is trying to figure out the group trust score.

[0133] For this subgroup, the trust scores were generated as shown in Figure 16. Figure 16 is a diagram showing an example of group trust scores based on Figure 15. In Figure 16, the first row is the item, the second row is when A is the host, the third and fourth rows are when B is the host, the fifth to seventh rows are when C is the host, and the eighth row is when E is the host. The third column is social trust, the fourth column is familial trust, the fifth column is professional trust, and the sixth column is the average value of columns three to five. The score calculation unit 206 selects a dominant reliability score (for example, a predetermined value (for example, 0.5) or more). In Fig. 16, the selected dominant score is highlighted as indicated by the symbol g101.

[0134] When the algorithm of this embodiment is used, the group trust score is as shown in Fig. 17. Fig. 17 is a diagram showing an example of a group trust score based on Fig. 16. In Fig. 17, the first row is the item, the second row is when A is the host, the third row is when B is the host, the fourth row is when C is the host, the fifth row is when D is the host, the fifth row is when E is the host, and the sixth row is the general case. The second column is social trust, the third column is familial trust, and the fourth column is professional trust. In FIG. 17, the dominant group confidence score is highlighted as g11, and the second dominant score is highlighted as g112.

[0135] In the above calculation, env scorevalue was set to (1.0,1.0,1.0) and the recommendation reliability score value was set to (0.0,0.0,0.0).

[0136] (Second evaluation) Suppose a large group of people gather at a colleague's home for his retirement party. Friends, family and other colleagues are also invited to the party. During the party, five specific individuals (denoted as A, B, C, D and E) are seated and discussing together with Robot 2. In this case, the group dynamics can be described as follows: D is a colleague who is the subject of a retirement party. B was D's mentor when he was a graduate student under B. D is A's boss and works at the same company. A and E are colleagues and close friends. A has heard D talk about B and knows that B is a respected scholar in a certain academic field. C is an old neighbour of B and has been invited to the party. C is also E's grandparent.

[0137] Robot 2 learns that such parties (social events) with many participants are held and uses the environment score env score = (1.3, 1.0, 1.0). Also, since Robot 2 can identify each individual and collect their privacy profile, it can obtain initial trust scores for the contacts as shown in Figure 18 (the dominant scores are highlighted). Figure 18 shows an example of initial trust scores in the second evaluation. In Figure 18, the first row is the item, the second and third rows are when A is the host, the fourth and fifth rows are when B is the host, the sixth and seventh rows are when C is the host, the eighth and nine rows are when C is the host, and the tenth and eleventh rows are when E is the host. The third column is social trust, the fourth column is familial trust, the fifth column is professional trust, and the sixth column is the average of columns three to five.

[0138] Using the algorithm of this embodiment, the group reliability scores shown in Fig. 19 are generated. Fig. 19 is a diagram showing an example of group reliability scores based on Fig. 18. In Fig. 19, the dominant group score is highlighted with the symbol g121, and the second most dominant score is highlighted with the symbol g122.

[0139] Based on these scores, there are some things that Robot 2 can determine. In Figure 19, among all group participants, there are higher trust scores related to social trust than work trust, which in turn is higher than family trust. Robot 2 may also infer that family-related topics are less appropriate unless C or E speaks as the host. Or Robot 2 may infer that work-related topics and prompts are more appropriate when D and A are speaking, but not when C is speaking.

[0140] The evaluations and evaluation results described with reference to FIGS. 14 to 19 are merely examples, and the present invention is not limited to these.

[0141] As described above, in this embodiment, a group trust score is used by the robot 2 to determine group trust dynamics between groups of individuals. This trust score is derived from various criteria of interpersonal trust relationships, environmental sensing of the location of the robot 2 and the user, recommendations given to specific individuals, etc.

[0142] In this embodiment, the robot 2 is also able to sense the environment it is in and determine the type and location of the gathering, and then attempts to identify individuals present using, for example, its modalities, sensors, influences, intelligence, etc. If an individual was successfully identified, Robot 2 retrieved that individual's privacy profile, including the trust scores for each of his or her contacts. Unidentified individuals were treated by Robot 2 as strangers to the group. Robot 2 then calculated a pairwise trust score for each possible pairing combination in the group and with respect to a particular host individual. Robot 2 then processed each pairwise score to generate a group trust score.

[0143] Thus, after processing the group trust scores, Robot 2 can better understand what the appropriate trust dynamics between individuals are. For example, if Robot 2 understands that there is trust between, say, two individuals, and understands the nature and context of that trust, it can prompt Robot 2 to react, perform and act in a more specific way when those individuals are present.

[0144] In the above example, the reliability score, group reliability score, and environment reliability score are shown as numerical values, but are not limited to this. The reliability score, group reliability score, and environment reliability score may be expressed as a word indicating a magnitude, such as "High", "Avg", "Above-Avg", "Low", etc., as shown in FIG.

[0145] In the above example, a robot is described as an example of an information processing device, but the information processing device is not limited to this. For example, the information processing device may be a mobile terminal, a tablet terminal, a smartphone, a stuffed animal, a mascot, a doll, an ornament, or the like that can be used for communication.

[0146] In the above example, the robot performs the processing, but the processing may be performed on the cloud.

[0147] In addition, a program for realizing all or part of the functions of the information processing system 1 in the present invention may be recorded in a computer-readable recording medium, and the program recorded in the recording medium may be read into a computer system and executed to perform all or part of the processing performed by the information processing system 1. Note that the term "computer system" as used herein includes hardware such as an OS and peripheral devices. The term "computer system" also includes a WWW system equipped with a homepage providing environment (or display environment). The term "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, and storage devices such as hard disks built into a computer system. The term "computer-readable recording medium" also refers to storage devices that hold a program for a certain period of time, such as volatile memory (RAM) inside a computer system that becomes a server or client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line.

[0148] The above program may also be transmitted from a computer system in which the program is stored in a storage device or the like to another computer system via a transmission medium, or by transmission waves in the transmission medium. Here, the "transmission medium" that transmits the program refers to a medium that has the function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. The above program may also be for realizing part of the above-mentioned functions. Furthermore, it may be a so-called difference file (difference program) that can realize the above-mentioned functions in combination with a program already recorded in the computer system.

[0149] The above describes the form for carrying out the present invention using an embodiment, but the present invention is not limited to such an embodiment, and various modifications and substitutions can be made within the scope that does not deviate from the gist of the present invention. [Explanation of symbols]

[0150] 1...information processing system, 2...robot (information processing device), 2-1...first robot, 2-2...second robot, 3...environmental sensor, 201, 201-1, 201-2...acquisition unit, 202, 202-1, 202-2...image processing unit, 203, 203-1, 203-2...audio processing unit, 204, 204-1, 204-2...motion processing unit, 205, 205-1, 205-2...recognition unit, 206, 20 6-1, 206-2... score calculation unit, 207, 207-1, 207-2... storage unit, 208, 208-1, 208-2... sensors, 209, 209-1, 209-2... actuators, 210, 210-1, 210-2... imaging unit, 212, 212-1, 212-2... sound collection unit, 214, 214-1, 214-2... policy processing unit, 301... imaging unit, 302... sound collection unit, 303... output unit

Claims

1. a score calculation unit that calculates a group reliability score representing a relationship between a plurality of users in a group including at least two users from a plurality of different reliability levels, and changes a handling level of user information based on the group reliability score; An information processing device comprising:

2. The plurality of different trust levels include three trust levels: social trust, family trust, and expert trust. The information processing device according to claim 1 .

3. The score calculation unit changes a weighting of each of the plurality of different trust levels according to an environment of the user. The information processing device according to claim 1 .

4. The score calculation unit is The role of each of the plurality of users is recognized, and a host role among the plurality of users is recognized based on the recognition result. If the host is recognized, a reliability score of a relationship from the other user to the host is calculated for each pair of the host and another user other than the host; calculating the group reliability score based on the reliability score calculated for each group; The information processing device according to claim 1 .

5. The score calculation unit is The role of each of the plurality of users is recognized, and a host role among the plurality of users is recognized based on the recognition result. If the host cannot be recognized, for each pair of users, calculate a reliability score of the relationship between the pair of users; calculating the group reliability score based on the reliability score calculated for each group; The information processing device according to claim 1 .

6. the handling level of the user information is one of: allowing all use of the user information, allowing some use of the user information, and not allowing any handling of the user information; The information processing device according to claim 1 .

7. One of the users is a person selected by the information processing device, The score calculation unit is calculating the trustworthiness scores of the selected people; calculating the group credibility score based on the calculated credibility scores of the selected persons; The information processing device according to claim 4.

8. An environmental sensor for detecting environmental information of an environment in which the plurality of users are present; The score calculation unit is An environment in which the plurality of users are present is estimated based on the environmental information detected by the environmental sensor, and an environment score for the environment in which the plurality of users are present is determined based on the estimation result; calculating the group trust score based on the determined environmental score; The information processing device according to claim 4.

9. a score calculation unit calculates a group reliability score representing a relationship between a plurality of users in a group including at least two users from a plurality of different reliability levels, and changes a handling level of the user information based on the group reliability score; Information processing methods.

10. The computer of the information processing device calculating a group reliability score representing a relationship between a plurality of users in a group including at least two users from a plurality of different reliability levels, and changing a handling level of user information based on the group reliability score; program.

Images