Behavioral planning program, information processing device, and behavioral planning method

Abstract

Develop action plans that are tailored to the user's emotions and circumstances. [Solution] The processing unit 12 inputs emotion information representing the user's emotions and situation information representing the user's situation into an action prediction model 15 that predicts an action plan according to the user's emotions and situation, and obtains an action prediction result from the action prediction model 15. The processing unit 12 also inputs the action prediction result into an emotion prediction model 16 that predicts the emotions that will result from executing the action plan, and obtains an emotion prediction result from the emotion prediction model 16. Furthermore, based on the action prediction result and the emotion prediction result, the processing unit 12 determines which action plan to propose to the user from among multiple action plans.

Description

[Technical Field]

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

[0002] Technologies have been proposed to predict people's future behavior and emotions. For example, a technology has been proposed that predicts future behavior and the emotional changes associated with that behavior based on past behavioral data of the target subject. Another technology has been proposed that defines multiple diffusion factors for a target subject based on surveys and news articles, and predicts the establishment of people's future behavior and habits from the time-series changes in the contribution of each diffusion factor. Furthermore, a technology has been proposed that predicts user behavior using user situations estimated from behavioral data about user behavior, user emotions estimated from motion data based on user movements, and experience model data in which emotions are associated with situations. In addition, a technology has been proposed that acquires an emotion value representing the user's emotions and notifies a supporter of a prompt to support the user if the emotion value is equal to a predetermined value representing the degree of anxiety or concern. Finally, a technology has been proposed that acquires physiological data of the user and data on the external environment, and if it is predicted that the user is heading towards an undesirable emotional state based on the acquired data, provides corrective measures. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] International Publication No. 2022 / 249483 [Patent Document 2] International Publication No. 2023 / 157063 [Patent Document 3] Japanese Patent Publication No. 2015-184764 [Patent Document 4] U.S. Patent Application Publication No. 2017 / 0046496 [Patent Document 5] U.S. Patent Application Publication No. 2018 / 0276345

Summary of the Invention

Problems to be Solved by the Invention

[0004] When formulating a user's action plan, if the user's emotions are not considered, it may not be possible to formulate an effective action plan. This is because emotions are one of the major factors influencing actions and can also be a factor determining the success or failure of actions.

[0005] In one aspect, this case aims to formulate an action plan according to the user's emotions and situation.

Means for Solving the Problems

[0006] In one proposal, emotion information representing the user's emotions and situation information representing the user's situation are input into an action prediction model that predicts an action plan according to the user's emotions and situation, and an action prediction result is obtained from the action prediction model. The action prediction result is input into an emotion prediction model that predicts the emotions that will occur by executing the action plan, and an emotion prediction result is obtained from the emotion prediction model. Based on the action prediction result and the emotion prediction result, an action plan planning program is provided that causes a computer to execute a process of determining the action plan to be proposed to the user among a plurality of action plans.

Effects of the Invention

[0007] According to one aspect, an action plan according to the user's emotions and situation can be formulated.

Brief Description of the Drawings

[0008] [Figure 1] It is a diagram showing an example of an action plan planning method according to the first embodiment. [Figure 2] It is a diagram showing a hardware example of an information processing apparatus according to the second embodiment. [Figure 3] It is a block diagram showing a functional example of an information processing apparatus. [Figure 4] This figure shows an example of an action plan list. [Figure 5] This figure shows an example of an emotion vector. [Figure 6] This figure shows an example of an action probability vector. [Figure 7] This figure shows an example of the first behavioral prediction model. [Figure 8] This figure shows an example of a second behavioral prediction model. [Figure 9] This is a flowchart showing an example of the processing procedure of the information processing device according to the second embodiment. [Modes for carrying out the invention]

[0009] The embodiments for carrying out the invention will be described below with reference to the drawings. [First Embodiment] The first embodiment is an action planning method for formulating action plans that are tailored to the user's emotions and circumstances. The following describes an action planning method for formulating action plans that improve the user's mental health, but it is not limited to this. For example, it can also be applied when formulating action plans for learning or skill acquisition, action plans for improving a lack of exercise, or action plans for overcoming various addictions.

[0010] Figure 1 shows an example of an action planning method according to the first embodiment. Figure 1 shows an information processing device 10 for implementing the action planning method. The information processing device 10 can implement the action planning method, for example, by executing an action planning program.

[0011] The information processing device 10 includes a storage unit 11 and a processing unit 12. The storage unit 11 is, for example, a memory or storage device of the information processing device 10. The processing unit 12 is, for example, a processor of the information processing device 10. The information processing device 10 may have multiple processors. The processor that executes one of the multiple processes performed by the information processing device 10 may be different from the processor that executes a different process.

[0012] The memory unit 11 stores, for example, an action planning program. The memory unit 11 also stores multiple action plans 11a. Action plans 11a are, for example, candidate actions to improve the user's mental health. Action plans 11a may include a variety of action plans with different degrees of feasibility and improvement effects, such as action plans that have a high improvement effect but are not feasible depending on the user's emotions and circumstances, and action plans that have a low improvement effect but are highly feasible.

[0013] The processing unit 12 executes the action planning method according to the processing procedure shown in the action planning program, for example. The processing procedure of the action planning method executed by the processing unit 12 is as follows:

[0014] The processing unit 12 inputs emotion information representing the user's emotions and situation information representing the user's situation into the behavior prediction model 15, and obtains behavior prediction results from the behavior prediction model 15. Hereinafter, this process will be referred to as the behavior prediction process.

[0015] As emotional information, for example, an emotional vector, which is a two-dimensional vector representing arousal and comfort levels, can be used. Russell's aerial emotional model is an emotional estimation model that estimates emotions based on arousal and comfort levels. Russell's aerial emotional model is a model in which each of a person's emotions is arranged in a ring on a two-dimensional plane with arousal on the vertical axis and comfort on the horizontal axis.

[0016] Furthermore, emotional information is not limited to emotional vectors, which are two-dimensional vectors as described above. For example, emotional vectors, which are three-dimensional vectors based on Millenson's three-dimensional model (a model that represents the intensity of basic emotions using three-dimensional vectors X, Y, and Z), can also be used as emotional information. Moreover, emotional information is not limited to the vector information described above.

[0017] The processing unit 12 can generate such emotional information based on, for example, the results of a survey conducted among users. The processing unit 12 may also receive such emotional information or survey results from a terminal device operated by the user via a network.

[0018] Situational information includes, for example, the date and time, the user's location, and the situation (whether it's a holiday or during working hours). This is because the possible action plans can vary depending on the date, time, location, and situation.

[0019] The processing unit 12 accepts the status information described above. The processing unit 12 may also receive the status information described above from a terminal device operated by the user via a network. The behavior prediction model 15 predicts an action plan based on the user's emotions and situation. The behavior prediction model 15 is, for example, a pre-trained machine learning model trained to output the probability of each of multiple action plans 11a based on the user's emotions and situation. The behavior prediction model 15 may also be a pre-trained machine learning model using a decision tree, trained to determine an action plan based on the user's emotions and situation.

[0020] Each machine learning model is trained, for example, based on the error between the model's processing results on training data (emotional and situational information) and the correct label (the executed action plan or its probability).

[0021] Figure 1 shows an example of behavioral prediction results based on emotional information representing the user's current emotions. In the example in Figure 1, the behavioral prediction results show that the probability of performing the action plan "Don't open the email immediately," which requires relatively little effort, is 60%, and the probability of performing the action plan "Meditate for 5 minutes," which requires relatively more effort, is 30%.

[0022] When a user's mental health is poor, they tend to be more likely to take action plans that require relatively little effort but have a low potential for improvement. Conversely, when a user's mental health is improving, they tend to be more likely to take action plans that require relatively more effort but have a high potential for improvement.

[0023] After the behavior prediction process, the processing unit 12 inputs the behavior prediction results into the emotion prediction model 16 and obtains the emotion prediction results from the emotion prediction model 16. Hereinafter, this process will be referred to as the emotion prediction process. The emotion prediction model 16 predicts the emotions that will result from executing the action plan. The emotion prediction model 16 outputs a predicted emotion result (emotion prediction result) of the emotions that will result from executing the action plan represented by the action prediction result.

[0024] The emotion prediction model 16 is a trained machine learning model that is trained to output emotion information representing the user's emotions when each of the multiple action plans 11a is executed. As emotion information, for example, emotion vectors, which are two-dimensional vectors representing arousal and comfort levels, can be used. The machine learning model is trained based on the error between the processing result of the trained model on the training data (action plans 11a) and the ground truth label (emotion information of the emotions obtained when action plan 11a is executed).

[0025] The processing unit 12 performs an action plan determination process to determine which action plan to propose to the user from among multiple action plans 11a, based on the action prediction results and the emotion prediction results. For example, if the processing unit 12 has obtained the action probability for each action plan 11a as an action prediction result, it determines which action plan to propose to the user as follows.

[0026] The processing unit 12 calculates an evaluation value for each action plan based on the respective action probabilities of the action plan 11a, the target emotion information representing the target emotion, and the emotion prediction results corresponding to each of the action plans 11a. Then, the processing unit 12 determines, according to the evaluation values of each action plan, the action plan to be proposed to the user among the action plans 11a.

[0027] The evaluation value (hereinafter referred to as the action evaluation score S t and called) can be expressed, for example, by the following formula (1).

[0028]

Equation

[0029] In formula (1), E T is the emotion information representing the target emotion (hereinafter referred to as the target emotion vector). E T is, for example, arbitrarily set by the user. E t,n is the emotion vector obtained by executing the nth action plan at time point t. The action evaluation score S t is proportional to the absolute value of the difference (Euclidean distance) between E T and E t,n . p' is the result of weighting the action probability p as shown in the following formula (2).

[0030]

Equation

[0031] To make the action evaluation score S t a better value (a smaller value) as the action probability p increases, p' becomes an exponentially larger value as the action probability p increases. The processing unit 12 determines, for example, the action plan with the smallest action evaluation score S t shown in formula (1) as the action plan to be proposed to the user.

[0032] The processing unit 12 outputs the determined action plan. The determined action plan may be displayed on a display device (not shown), for example, or transmitted via a network to a terminal device operated by the user.

[0033] The processing unit 12 may repeat the behavior prediction process, in which the emotion prediction result and situational information representing the future situation based on the user's schedule are input to the behavior prediction model 15, and the emotion prediction process a predetermined number of times. If one behavior prediction process and emotion prediction process predicts a day's action plan and the emotions that will result from executing that action plan, then a predetermined number of action plans and emotions will be predicted.

[0034] As behavioral prediction and emotion prediction processes are repeated, the predicted emotions and action plans can change. For example, even if a user initially implements an action plan with a relatively low impact on improving their mental health, their emotions may gradually become more positive, increasing the likelihood that they will be able to implement an action plan that requires more effort but has a higher impact on improvement.

[0035] Figure 1 shows an example of behavioral prediction results based on predicted emotions after behavioral prediction processing and emotion prediction processing have been repeated a predetermined number of times. In the example in Figure 1, the behavioral prediction results show that the probability of performing the relatively low-effort action plan, "Don't open the email immediately," is 50%, and the probability of performing the relatively high-effort action plan, "Meditate for 5 minutes," is 60%.

[0036] Furthermore, if the behavioral probability for each behavioral plan 11a is obtained as a result of the behavioral prediction, the processing unit 12 will, for example, select the behavioral evaluation score S shown in equation (1) from among the behavioral plans 11a. t Alternatively, the system may narrow down the options to N action plans in ascending order of the value. Then, the processing unit 12 inputs the emotion prediction result corresponding to each of the N action plans into the action prediction model 15 for the next action prediction process. This reduces the amount of computation required.

[0037] In the action plan determination process, the processing unit 12 determines an action plan based on the action prediction results for a predetermined number of times and the emotion prediction results for a predetermined number of times. In determining the action plan, the processing unit 12 uses an action evaluation score S as shown in equation (1). t The action plan (a combination of action plans for a predetermined number of times) that minimizes the sum of the values ​​may be determined. t The sum S total This can be expressed by the following equation (3).

[0038]

number

[0039] As described above, according to the behavior planning method of the first embodiment, the processing unit 12 inputs emotion information representing the user's emotions and situation information representing the user's situation into the behavior prediction model 15 and obtains behavior prediction results from the behavior prediction model 15. The behavior prediction model 15 is a model that predicts an action plan according to the user's emotions and situation. The processing unit 12 also inputs the behavior prediction results into the emotion prediction model 16 and obtains emotion prediction results from the emotion prediction model 16. The emotion prediction model 16 is a model that predicts the emotions that will result from executing the action plan. Furthermore, the processing unit 12 performs an action plan determination process to determine an action plan to propose to the user based on the behavior prediction results and emotion prediction results. This makes it possible to formulate an action plan that is appropriate to the user's situation and emotions.

[0040] Furthermore, according to the behavior planning method of the first embodiment, the processing unit 12 repeats the behavior prediction process, in which the emotion prediction result and situation information representing the future situation based on the user's plans are input to the behavior prediction model 15, and the emotion prediction process a predetermined number of times. Then, in the behavior plan determination process, the processing unit 12 determines the behavior plan based on the behavior prediction results and emotion prediction results from a predetermined number of times.

[0041] Improving and maintaining emotions and behaviors can be difficult with short-term approaches. This is because actions that produce significant emotional changes in the short term are inherently difficult to implement. In particular, it is difficult to encourage rapid improvement in individuals with poor mental health, but as their mental health gradually improves, the behaviors they can implement will also change.

[0042] If emotions are not considered when formulating action plans, it is impossible to anticipate how the action plan will change in response to future emotional fluctuations. Therefore, when predicting and proposing the long-term improvement effect on the user's condition, it may only be possible to propose a limited range of options. In particular, when proposing action plans to improve mental health, it is difficult to propose highly effective action plans to people whose mental health is currently deteriorating. However, if only action plans with low improvement effect are proposed, the long-term improvement effect will remain low. Furthermore, because the long-term improvement effect cannot be demonstrated, there is a problem in that users may not be able to accept and act upon the plan.

[0043] According to the behavioral planning method of the first embodiment described above, it is possible to predict gradual changes in emotions and behavioral plans, thereby formulating behavioral plans that are highly feasible in the long term and have a high effect in improving the user's condition. Furthermore, it becomes possible to present long-term effects to the user, which is expected to increase their acceptance, improve their motivation, and lead to action.

[0044] For example, daily self-care is important for improving and maintaining mental health. However, the self-care that can be performed varies depending on the user's mental health status. According to the action planning method of the first embodiment, it is possible to propose an action plan that shows simple self-care in the initial stage, and then gradually move to an action plan that shows self-care with a greater improvement effect.

[0045] [Second Embodiment] Next, a second embodiment will be described. Figure 2 shows an example of the hardware of an information processing device according to the second embodiment. Figure 2 shows an information processing device 20 for implementing the action planning method according to the second embodiment. The information processing device 20 can implement the action planning method, for example, by executing an action planning program. The information processing device 20 may also be called a computer. The information processing device 20 may be a client device or a server device.

[0046] The information processing device 20 includes a processor 21, RAM (Random Access Memory) 22, HDD (Hard Disk Drive) 23, GPU (Graphics Processing Unit) 24, input interface 25, media reader 26, and communication interface 27. These units are connected to a bus. The processor 21 corresponds to the processing unit 12 of the first embodiment. The RAM 22 or HDD 23 corresponds to the storage unit 11 of the first embodiment.

[0047] The processor 21 is a processor such as a GPU or CPU (Central Processing Unit) that includes arithmetic circuits for executing program instructions. The processor 21 loads at least a portion of the program and data stored in the HDD 23 into the RAM 22 and executes the program. The processor 21 may have multiple processor cores. The information processing device 20 may also have multiple processors. The processor that executes one of the multiple processes performed by the information processing device 20 may be different from the processor that executes a different process from the multiple processes. The processor may also be called a processor circuitry. A collection of multiple processors (multiprocessor) may also be called a "processor".

[0048] RAM22 is a volatile semiconductor memory that temporarily stores programs executed by the processor 21 and data used by the processor 21 for calculations. The information processing device 20 may also be equipped with other types of memory besides RAM22, and may be equipped with multiple types of memory.

[0049] HDD23 is a non-volatile storage device that stores software programs such as the OS (Operating System), middleware, and application software, as well as data. The programs include an action planning program that causes the information processing device 20 to execute an action planning process that formulates an action plan according to the user's emotions and situation. The information processing device 20 may also be equipped with other types of storage devices such as flash memory or SSD (Solid State Drive), and may be equipped with multiple non-volatile storage devices.

[0050] The GPU 24 outputs an image to the display 24a connected to the information processing unit 20, according to instructions from the processor 21. The display 24a can be a CRT (Cathode Ray Tube) display, an LCD (Liquid Crystal Display), a PDP (Plasma Display Panel), an OEL (Organic Electro-Luminescence) display, or the like.

[0051] The input interface 25 acquires input signals from input devices 25a connected to the information processing device 20 and outputs them to the processor 21. Input devices 25a can include pointing devices such as mice, touch panels, touchpads, and trackballs, as well as keyboards, remote controllers, and button switches. Furthermore, multiple types of input devices may be connected to the information processing device 20.

[0052] The media reader 26 is a reading device that reads programs and data recorded on the recording medium 26a. Examples of recording media 26a include magnetic disks, optical disks, magneto-optical disks (MO), and semiconductor memory. Magnetic disks include flexible disks (FD) and HDDs. Optical disks include CDs (Compact Discs) and DVDs (Digital Versatile Discs).

[0053] The media reader 26 copies programs and data read from the recording medium 26a to other recording media such as RAM 22 or HDD 23. The read programs are executed by the processor 21, for example. The recording medium 26a may be a portable recording medium and may be used for distributing programs and data. The recording medium 26a and HDD 23 may also be referred to as computer-readable recording media.

[0054] The communication interface 27 is connected to the network 27a and communicates with other information processing devices via the network 27a. In the example in Figure 2, the communication interface 27 communicates with the terminal device 40 via the network 27a. The terminal device 40 may be a PC (Personal Computer), smartphone, tablet, etc. The terminal device 40 is, for example, a terminal device operated by a user requesting action plan suggestions. The communication interface 27 may be a wired communication interface connected by a cable to a communication device such as a switch, or it may be a wireless communication interface connected by a wireless link to a base station.

[0055] Next, the functions of the information processing device 20 will be explained. Figure 3 is a block diagram showing an example of the functions of an information processing device. The information processing device 20 includes an input unit 30, an action plan list creation unit 31, an action plan list storage unit 32, an emotion / personal characteristic vector generation unit 33, an action prediction unit 34, an emotion prediction unit 35, a proposed plan determination unit 36, and an output unit 37. These units enable the same functions as the storage unit 11 and processing unit 12 shown in Figure 1.

[0056] The input unit 30, the action plan list creation unit 31, the emotion / personal characteristic vector generation unit 33, the action prediction unit 34, the emotion prediction unit 35, the proposed plan determination unit 36, and the output unit 37 can be implemented, for example, using program modules executed by the processor 21. The action plan list storage unit 32 is implemented using memory areas allocated in RAM 22 or HDD 23.

[0057] The input unit 30 acquires the results of a questionnaire administered to the user and TPO (Time Place Occasion) information. For example, the input unit 30 acquires the results of a questionnaire that includes multiple questions that reflect the user's current level of alertness and comfort. The level of alertness and comfort is information used to generate the aforementioned two-dimensional vector. The response results may also include answers to questions that investigate the emotions the user is aiming for.

[0058] Furthermore, the survey results may include answers to questions designed to generate personal trait information (hereinafter referred to as personal trait vectors) that represent the user's personal characteristics. As a method for classifying a user's personal characteristics, the Big Five personality model can be used, which divides human personality traits into five basic characteristics: extraversion, agreeableness, conscientiousness, neuroticism, and openness. A questionnaire that evaluates personal characteristics using the Big Five may include questions to explore the degree of extraversion, such as "I think I am active and extroverted," or "I think I am reserved and quiet." Similarly, questions to explore the degree of agreeableness may include questions such as "I think I am easily annoyed by others and prone to conflict," or "I think I am a considerate and kind person." Questions exploring the degree of the other three traits may also be included in the questionnaire.

[0059] Furthermore, in order to more appropriately evaluate individual characteristics, the questionnaire may include questions about statistical information such as gender, age, and country of birth, as well as questions about physical information such as height and weight, in addition to the Big Five personality traits mentioned above.

[0060] TPO information corresponds to the aforementioned situational information. TPO information includes the date and time, the user's location, and the situation (such as whether it is a holiday or during working hours). If the information processing device 20 predicts changes in the user's emotions and behavioral plans over several days, the TPO information also includes information representing the future situation based on the user's schedule over several days.

[0061] Furthermore, the input unit 30 acquires multiple action plans that can be proposed to the user. Multiple action plans may be entered by the operator of the information processing device 20 through the operation of the input device 25a, or they may be received from other computers via the network 27a. Survey responses and TPO information may be received, for example, from a terminal device 40 operated by the user via the network 27a.

[0062] The action plan list creation unit 31 creates an action plan list based on the action plan acquired by the input unit 30. Figure 4 shows an example of an action plan list. The action plan list 33a shown in Figure 4 includes multiple action plans that are candidates for actions to improve the user's mental health.

[0063] Action plan list 33a includes a variety of action plans with varying degrees of feasibility and improvement effects, such as action plans that have a high improvement effect but are less feasible depending on the user's emotions and circumstances, and action plans that have a low improvement effect but are highly feasible.

[0064] The input unit 30 may also obtain the number of repetitions when the behavior prediction process and emotion prediction process are repeated a predetermined number of times. The action plan list storage unit 32 stores the action plan list created by the action plan list creation unit 31.

[0065] The emotion / personal trait vector generation unit 33 generates the current emotion vector E0 and the target emotion vector E based on the questionnaire responses obtained by the input unit 30. T This generates an individual characteristic vector P.

[0066] Figure 5 shows an example of an emotion vector. In the example in Figure 5, on a two-dimensional plane with arousal level on the vertical axis and comfort level on the horizontal axis, the emotion vector represents each emotion expressed by arousal level and comfort level as a probability. Such emotion vectors are referred to as the current emotion vector E0 and the target emotion vector E T It can be used as such.

[0067] The individual trait vector P can be represented, for example, by the degree of each of the five basic traits mentioned above: extraversion, agreeableness, conscientiousness, neuroticism, and openness. The behavior prediction unit 34 predicts an action plan that is appropriate to the user's emotions and situation from among multiple action plans included in the action plan list stored in the action plan list storage unit 32. The behavior prediction unit 34 uses a trained machine learning model, which is the behavior prediction model, to input the current emotion vector E0 or the emotion prediction result E described later, which is the emotion vector E t,n Then, TPO information and personal characteristic vector P are input. The behavior prediction unit 34 then obtains candidate behavior plans as behavior prediction results from the behavior prediction model.

[0068] Below are two examples of behavior prediction models. The first behavior prediction model outputs behavior probability vectors as candidate behavior plans, representing the probability of each behavior plan depending on the user's emotions and situation.

[0069] Figure 6 shows an example of an action probability vector. In the example in Figure 6, the user's action probability is shown for each of the action plans A through P. The action probabilities for action plans D, K, and L are higher than those for the other action plans.

[0070] Figure 7 shows an example of the first behavior prediction model. The behavior prediction model in Figure 7 is a trained neural network model 34a. As the neural network model 34a, for example, a CNN (Convolutional Neural Network), an RNN (Recurrent Neural Network), or a Transformer can be used.

[0071] The neural network model 34a is trained to output behavioral probability vectors corresponding to the user's emotions, situation, and personal characteristics. The neural network model 34a outputs either the current emotion vector E0 or the emotion prediction result E described later. t,n Individual characteristic vectors P and TPO information are input. The neural network model 34a then outputs an action probability vector.

[0072] The emotion vector E at each point in time for which emotion prediction results were obtained. t,n By inputting TPO information at each point in time into the neural network model 34a, time-series data of action probability vectors can be obtained.

[0073] Figure 8 shows an example of a second behavior prediction model. The behavior prediction model in Figure 8 is a trained decision tree 34b. Random forests, gradient boosting decision trees, etc., can be used as the decision tree 34b.

[0074] In the example of decision tree 34b in Figure 8, branching is first performed based on the level of diligence obtained from the individual trait vector P. Furthermore, emotion vectors E0, E t,n Branching occurs depending on whether the arousal level obtained is 0 or greater. Then, the emotional vectors E0,E t,nBased on the comfort level obtained, or TPO information, branching occurs to determine candidate action plans. In the example in Figure 8, the candidate action plans determined differ depending on whether the comfort level is 3 or higher, whether the action plan is to be determined for 25 days or more from the present (Day ≥ 25), whether the user is at home or not, and whether the user is with someone or not.

[0075] The emotion prediction unit 35 inputs the behavior prediction result into an emotion prediction model, which is a trained machine learning model, and receives an emotion vector E from the emotion prediction model as the emotion prediction result. t,n The emotion prediction unit 35 is implemented using an emotion prediction model, which is a trained machine learning model. Emotion vector E t,n This represents the emotion that arises when the nth action plan in the action plan list is executed at time t (the tth iteration).

[0076] For emotion prediction models, similar to behavior prediction models, neural network models such as CNNs, RNNs, and Transformers, or decision trees such as random forests and gradient boosting decision trees can be used.

[0077] The proposed plan determination unit 36 ​​performs an action plan determination process to determine an action plan to propose to the user based on the action prediction results and the emotion prediction results. If the action prediction results include the action probability for each action plan, the proposed plan determination unit 36 ​​uses p' obtained by weighting the action probability according to equation (2) and the emotion vector E t,n And, target emotion vector E T Therefore, the behavioral evaluation score S in equation (1) t The proposed plan determination unit 36 ​​then calculates the behavioral evaluation score S as shown in equation (1). t The action plan that minimizes [the specified factor] is selected as the action plan to propose to the user.

[0078] When predicting an action plan spanning multiple days, or when predicting a predetermined number of action plans based on the emotion prediction results, the proposed plan determination unit 36 ​​determines a predetermined number of action plan combinations (e.g., multiple days). In this case, the proposed plan determination unit 36 ​​determines, for example, the action evaluation score S represented by equation (3). t The sum S total Determine the combination of action plans for a predetermined number of times that minimizes [the specified value].

[0079] The output unit 37 outputs the determined action plan. The output unit 37 transmits the determined action plan to, for example, a terminal device 40 operated by the user via the network 27a. Alternatively, the output unit 37 may display the determined action plan on the display 24a, or store the determined action plan in a storage device such as the HDD 23.

[0080] Next, the processing procedure of the information processing device 20 will be explained. Figure 9 is a flowchart showing an example of the processing procedure of the information processing device according to the second embodiment. The process shown in Figure 9 will be described below in accordance with the step numbers.

[0081] [Step S10] Information is acquired by the input unit 30. The information acquired by the input unit 30 includes the results of a questionnaire targeting the user, TPO information, and multiple action plans that can be proposed to the user.

[0082] [Step S11] The action plan list creation unit 31 creates an action plan list. The created action plan list is stored in the action plan list storage unit 32. [Step S12] The emotion / personal trait vector generation unit 33 generates the current emotion vector E0 and the target emotion vector E based on the questionnaire response results obtained by the input unit 30. T This generates an individual characteristic vector P.

[0083] [Step S13] The behavior prediction unit 34 performs behavior prediction processing. In the behavior prediction processing, the current emotion vector E0 or the emotion prediction result emotion vector E0 is applied to the behavior prediction model. t,n Then, TPO information and individual characteristic vector P are input. The behavior prediction model then retrieves candidate behavior plans as behavior prediction results.

[0084] [Step S14] The emotion prediction unit 35 performs emotion prediction processing. In emotion prediction processing, the behavior prediction result is input to the emotion prediction model. Then, the emotion prediction model outputs an emotion vector E as the emotion prediction result. t,n This is obtained.

[0085] [Step S15] If the proposed plan determination unit 36 ​​includes the action probability of each action plan in the action prediction result, it uses p' obtained by weighting the action probability according to equation (2) and the emotion vector E t,n And, target emotion vector E T Therefore, the behavioral evaluation score S in equation (1) t Calculate.

[0086] [Step S16] The proposed plan determination unit 36 ​​selects an action evaluation score S from all action plans included in the action plan list. t Select N action plans in ascending order. [Step S17] The proposed plan determination unit 36 ​​determines whether it has completed predicting a predetermined number of action plans (hereinafter referred to as D times). If the proposed plan determination unit 36 ​​determines that it has completed predicting D times the action plans, it performs the process in step S18. If the proposed plan determination unit 36 ​​determines that it has not completed predicting D times the action plans, the process from step S13 is repeated. When N action plans have been selected, in the process of step S13, the action prediction model receives the emotion vector E, which is the emotion prediction result for the N selected action plans. t,n Then, TPO information and personal characteristic vector P are input. And N emotion vectors E t,n Candidate action plans for each of these are obtained from the behavioral prediction model.

[0087] [Step S18] The proposed plan determination unit 36 ​​is N D For each combination of street action plans, the action evaluation score S t The sum S total Calculate. [Step S19] The proposed plan determination unit 36 ​​is N D Among the combinations of action plans, the action evaluation score S t The sum S total Determine the combination of D action plans that minimizes the problem.

[0088] [Step S20] The output unit 37 outputs the decided action plan. Note that the above processing procedure is just one example, and the order of processing may be changed as appropriate. Also, if one action plan is determined in a single action prediction process, as in the case of using decision tree 34b as shown in Figure 8, steps S15, S16, and S18 do not need to be performed.

[0089] According to the behavioral planning method of the second embodiment described above, it is possible to predict gradual changes in emotions and behavioral plans, thereby formulating behavioral plans that are highly feasible in the long term and have a high effect in improving the user's condition. Furthermore, it becomes possible to present long-term effects to the user, which is expected to increase their acceptance, improve their motivation, and lead to action.

[0090] Furthermore, the information processing device 20 selects the action evaluation score S from among multiple action plans. t N action plans are selected in ascending order, and the emotion prediction results corresponding to each of the N action plans are input into the action prediction model for the next action prediction process. This reduces the amount of computation required.

[0091] Furthermore, by using personal characteristic information (personal characteristic vectors) to perform behavioral prediction processing, it becomes possible to propose behavioral plans that are more suitable for the user. Furthermore, the information processing device 20 may send commands to the terminal device 40, which the user operates according to the action plan proposed to the user, to put the power to sleep mode or to display a message when the user attempts to perform a specific action (for example, checking email).

[0092] Although embodiments have been illustrated above, the configurations of each part shown in the embodiments can be replaced with others having similar functions. Furthermore, other arbitrary components or processes may be added. Moreover, any two or more configurations (features) from the embodiments described above may be combined. [Explanation of Symbols]

[0093] 10 Information Processing Devices 11 Storage section 11a Action Plan 12 Processing Units 15 Behavioral Prediction Models 16. Emotion Prediction Models

Claims

1. A behavioral prediction model that predicts action plans according to the user's emotions and situation is input with emotion information representing the user's emotions and situation information representing the user's situation, and the behavioral prediction result is obtained from the behavioral prediction model. The behavioral prediction results are input into an emotion prediction model that predicts the emotions that will result from executing the aforementioned action plan, and the emotion prediction results are obtained from the emotion prediction model. Based on the behavior prediction results and the emotion prediction results, the system determines which of the multiple behavior plans to propose to the user. An action planning program that directs a computer to perform a task.

2. The process of inputting the emotion prediction result and the situation information representing the future situation based on the user's plans into the behavior prediction model to obtain the behavior prediction result, and the process of inputting the behavior prediction result into the emotion prediction model to obtain the emotion prediction result, is repeated a predetermined number of times. In the process of determining the action plan, the action plan is determined for a predetermined number of times based on the action prediction results for a predetermined number of times and the emotion prediction results for a predetermined number of times. The action planning program according to claim 1, which causes the computer to perform the processing.

3. The behavior planning program according to claim 1, wherein the behavior prediction model is a trained machine learning model trained to output the probability of each of the plurality of behavior plans taking action according to the user's emotions and circumstances.

4. The process for determining the aforementioned action plan is: Based on the action probability of each of the multiple action plans, the target emotion information representing the target emotion, and the emotion prediction results corresponding to each of the multiple action plans, an evaluation value for each of the multiple action plans is calculated. Based on the evaluation value, the action plan to propose to the user is determined from among the multiple action plans. The action planning program according to claim 3, including processing.

5. The behavioral planning program according to claim 1, wherein the behavioral prediction model is a pre-trained machine learning model using a decision tree that is trained to determine a behavioral plan from the plurality of behavioral plans according to the user's emotions and situation.

6. The aforementioned behavior prediction model is further input with personal characteristic information representing the user's personal characteristics, The behavioral prediction model predicts the behavioral plan according to the user's emotions, situation, and personal characteristics. The behavioral planning program according to claim 1.

7. A processing unit inputs emotional information representing the user's emotions and situational information representing the user's situation into an action prediction model that predicts an action plan according to the user's emotions and situation, obtains an action prediction result from the action prediction model, inputs the action prediction result into an emotion prediction model that predicts the emotions that will result from executing the action plan, obtains an emotion prediction result from the emotion prediction model, and determines the action plan to propose to the user from among a plurality of action plans based on the action prediction result and the emotion prediction result. An information processing device having

8. A behavioral prediction model that predicts action plans according to the user's emotions and situation is input with emotion information representing the user's emotions and situation information representing the user's situation, and the behavioral prediction result is obtained from the behavioral prediction model. The behavioral prediction results are input into an emotion prediction model that predicts the emotions that will result from executing the aforementioned action plan, and the emotion prediction results are obtained from the emotion prediction model. Based on the behavior prediction results and the emotion prediction results, the system determines which of the multiple behavior plans to propose to the user. A method of planning actions that a computer will execute.

Images