Method and system for maintaining and automatically optimizing personality consistency of accompanying AI

Abstract

The invention relates to a personality consistency maintenance and automatic optimization method and system for an accompanying AI. The method comprises the following steps: adding a personality agent for the accompanying AI; constructing a multi-level personality vector model in the personality agent, and constructing a personality vector through the multi-level personality vector model; adding a personality consistency protocol into the personality agent; adding a personality coordination and conflict arbitration module into the personality agent, wherein the personality coordination and conflict arbitration module compares and corrects output results of different task agents; a personality automatic optimization mechanism is added into a personality agent, the personality automatic optimization mechanism continuously collects user feedback in the operation process, the feedback is input into a personality vector optimizer through a self-adaptive feedback loop, and a multi-level personality vector model is driven to conduct dynamic iteration updating. According to the method, personality consistency can be realized in a multi-agent collaborative scene, and personality expression can be optimized through interaction with the user.

Description

Technical Field

[0001] This invention relates to the field of human-computer interaction technology, and in particular to a method and system for maintaining and automatically optimizing the personality consistency of companion AI. Background Technology

[0002] Existing companion-oriented AI systems are primarily based on a single large model or single-agent architecture. To simultaneously provide emotional companionship, medical companionship, professional advice (such as legal or financial advice), long-term memory, real-time perception, and tool operation, a single model is either too large and difficult to train, or performs poorly in certain specialized scenarios. Multi-agent systems can address the problems of single-agent systems by having multiple specialized single agents work collaboratively. However, existing multi-agent systems, lacking a unified personality coordination mechanism, are prone to the following issues:

[0003] Poor personality consistency: Although multi-agent systems can complete complex tasks through the collaboration of multiple professional agents, the lack of a unified mechanism in terms of tone, values ​​and behavior among the agents leads to a fragmented user experience and can easily create a sense of "personality split".

[0004] Personality template-based and rigid: Existing companion AI relies heavily on fixed template settings, lacking the ability for dynamic evolution of personality traits. As the interaction context becomes longer, simply relying on templates makes it difficult to achieve personalized optimization, resulting in a monotonous and formulaic companionship experience.

[0005] Lack of adaptive optimization capabilities: The existing system cannot dynamically adjust personality performance based on long-term user feedback, resulting in a stagnant companionship experience and difficulty in meeting users' personalized and growth-oriented needs.

[0006] Therefore, it is necessary to provide a method and system for maintaining and automatically optimizing the personality consistency of companion AI, which can enable personality consistency in multi-agent collaborative scenarios and optimize personality performance through interaction with users. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for maintaining and automatically optimizing the personality consistency of companion AI, which enables personality consistency in multi-agent collaborative scenarios and optimizes personality performance through interaction with users.

[0008] To address the problems existing in the prior art, this invention provides a method for maintaining and automatically optimizing the personality consistency of companion AI, comprising the following steps:

[0009] Add a personal intelligent agent to companion AI;

[0010] A multi-level personality vector model is constructed in the personality intelligent agent, and personality vectors are constructed through the multi-level personality vector model; the multi-level personality vector model includes a core personality layer, a task-adaptive personality layer and a scene dynamic layer;

[0011] A personality consistency protocol is incorporated into the personality agent. The personality consistency protocol includes constraint format, consistency verification PCP definition rules, conflict resolution, and dynamic control.

[0012] A personality coordination and conflict arbitration module is added to the personality intelligence agent. The personality coordination and conflict arbitration module compares and corrects the output results of different task intelligence agents.

[0013] An automatic personality optimization mechanism is added to the personality intelligent agent. During operation, the automatic personality optimization mechanism continuously collects user feedback. The feedback is input into the personality vector optimizer through an adaptive feedback loop, driving the multi-level personality vector model to perform dynamic iterative updates.

[0014] Optionally, in the method for maintaining and automatically optimizing the personality consistency of the companion AI, a multi-level personality vector model is constructed based on user interaction history, semantic features, and behavioral preferences. The personality vector is constructed through the personality vector model, and the personality vector is distributed to each task agent as a global control condition. The personality constraints are automatically embedded in each task agent when performing task reasoning and decision-making.

[0015] Optionally, in the method for maintaining and automatically optimizing the personality consistency of the companion AI, the multi-level personality vector model includes the following three layers:

[0016] Core personality layer: long-term stable traits, including values ​​and tone of voice;

[0017] Task-adaptive personality layer: Adjustable characteristics related to specific tasks;

[0018] Scene dynamic layer: Personality parameters that are adjusted in real time according to the environment and situation.

[0019] Optionally, in the method for maintaining and automatically optimizing the personality consistency of the companion AI, the personality consistency protocol is called Persona Consistency Protocol, or PCP for short.

[0020] Optionally, in the method for maintaining and automatically optimizing the personality consistency of the companion AI, the personality consistency protocol includes:

[0021] Constraint format: Specifies the way personality vectors are represented and transmitted between task agents, and ensures that different task agents receive the same vector with consistent meaning;

[0022] Consistency check (PCP) definition rules: Ensure that the results generated by each task agent conform to personality constraints;

[0023] Conflict resolution: When two task agents have a style conflict while handling the same task, PCP provides a priority arbitration strategy.

[0024] Dynamic adjustment: When switching scenarios, the method of shifting personality parameters is specified, while ensuring that core personality traits remain unchanged.

[0025] Optionally, in the method for maintaining and automatically optimizing the personality consistency of the companion AI, the representation and transmission method are standardized into a unified JSON or embedding.

[0026] Optionally, in the method for maintaining and automatically optimizing the personality consistency of the companion AI, user feedback includes explicit feedback and implicit feedback. The explicit feedback includes rating, modification, correction, and selection, while the implicit feedback includes interaction frequency, dwell time, emotional fluctuations, and emotional reactions.

[0027] Optionally, in the method for maintaining and automatically optimizing the personality consistency of the companion AI, the automatic personality optimization mechanism includes:

[0028] Feedback collection: Simultaneously collect both explicit and implicit feedback from users as input signals for personality vector optimization;

[0029] Reward modeling: Task success rate, human-computer interaction quality and personality consistency are combined to calculate the reward signal, and a penalty term is introduced to suppress deviations that do not conform to the personality consistency protocol;

[0030] Layered updates: Different update rates are set according to the levels of the personality vector to ensure the stability of core personality traits, while allowing task- and scenario-related parameters to adapt quickly;

[0031] Consistency constraints: Hard and soft constraints of the personality consistency protocol are introduced during the optimization process to avoid abnormal drift of personality vectors and ensure the uniformity of output style of multiple agents;

[0032] Steady-state control: By using confidence weighting, anomaly monitoring and snapshot rollback mechanisms, the impact of single noise feedback is reduced, and the system rolls back to a steady state when it deviates from a preset threshold.

[0033] This invention provides a system for maintaining and automatically optimizing the personality consistency of companion AI. The system is established by employing the aforementioned method for maintaining and automatically optimizing the personality consistency of companion AI.

[0034] Compared with the prior art, the present invention has the following advantages:

[0035] (1) Ensuring consistency of personality among multiple agents: Through personality vector modeling and the Personality Consistency Protocol (PCP), each task agent can share the same set of personality traits when performing different tasks, thereby avoiding inconsistencies in output tone, values, and behavioral styles. When interacting with the system, users will continuously experience a consistent personality, significantly improving the naturalness and immersion of human-computer interaction.

[0036] (2) Improve the efficiency of multi-agent collaboration: The introduction of a personality coordination and conflict arbitration module enables different task agents to quickly resolve tone and style conflicts when handling complex tasks, reducing redundant calculations and repetitive outputs. The overall operating efficiency of the system and the stability of the user experience are both improved.

[0037] (3) Support for personalization and long-term adaptation: Through the adaptive optimization mechanism of personality vector, the system can automatically adjust personality parameters based on explicit and implicit feedback during long-term use by users. This process ensures the stability of core personality traits and allows for rapid adaptation between the task layer and the scenario layer, thereby achieving personalization and continuous growth of personality.

[0038] (4) Avoid personality drift and abnormal output: ensure the long-term stability of the personality performance of the companion AI and avoid personality mutation or style abnormality.

[0039] (5) Enhanced Application Value: The method provided by this invention can be widely applied to scenarios such as intelligent assistants, virtual companions, educational robots, and medical care. The ability to maintain consistency in personality and dynamically optimize in different scenarios can significantly enhance users' trust and long-term user stickiness. Attached Figure Description

[0040] Figure 1 A flowchart illustrating the personality consistency maintenance and automatic optimization method provided in this embodiment of the invention. Detailed Implementation

[0041] The specific embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. The advantages and features of the present invention will become clearer from the following description. It should be noted that the drawings are all in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the present invention.

[0042] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0043] In the following, if the methods described herein include a series of steps, the order of these steps presented herein is not necessarily the only order in which these steps can be performed, and some of the steps described may be omitted and / or some other steps not described herein may be added to the method.

[0044] Existing companion-oriented AI systems are primarily based on a single large model or single-agent architecture. To simultaneously provide emotional support, medical assistance, professional advice (such as legal or financial advice), long-term memory, real-time perception, and tool operation, a single model is either too large and difficult to train, or performs poorly in certain specialized scenarios. Multi-agent systems can address the problems of single-agent systems by having multiple specialized single agents work collaboratively. However, existing multi-agent systems lack a unified personality coordination mechanism, leading to issues such as poor personality consistency, personality template-based rigidity, and a lack of adaptive optimization capabilities.

[0045] To address the problems existing in the prior art, this invention provides a method for maintaining and automatically optimizing personality consistency in companion-type AI (Artificial Intelligence), such as... Figure 1 The method includes the following steps:

[0046] S1: Add a personal intelligent agent to companion-type AI;

[0047] S2: Construct a multi-level personality vector model in the personality agent, and construct personality vectors through the multi-level personality vector model; the multi-level personality vector model includes a core personality layer, a task-adaptive personality layer, and a scene dynamic layer;

[0048] S3: To prevent personality bias in multi-task agents operating independently, this invention incorporates a Persona Consistency Protocol (PCP) into the personality agents. PCP is a set of rules and constraints that guide multiple agents to maintain consistency when using personality vectors. PCP specifies constraints for each task agent regarding language style, emotional tone, and value expression. All task agents must adhere to PCP when generating responses, thus ensuring personality uniformity. The Persona Consistency Protocol includes constraint format, PCP definition rules for consistency verification, conflict resolution, and dynamic adjustment.

[0049] S4: Add a personality coordination and conflict arbitration module to the personality intelligence agent. The personality coordination and conflict arbitration module compares and corrects the output results of different task intelligence agents. When personality conflict or style inconsistency is detected, the coordination module arbitrates according to priority, task weight and user preference, and unifies the results when necessary.

[0050] S5: An automatic personality optimization mechanism is added to the personality intelligence agent. This mechanism continuously collects user feedback during operation, and this feedback is input into the personality vector optimizer through an adaptive feedback loop, driving the multi-level personality vector model to dynamically iterate and update. This allows the companion AI to continuously optimize according to the user's long-term habits and changing needs, reflecting growth and personalization.

[0051] The purpose of this invention is to provide a method for maintaining personality consistency and automatically optimizing personality in multi-agent companion AI, addressing the following technical problems: ensuring consistent personality performance in multi-agent systems during task execution and avoiding personality fragmentation. It also enables dynamic self-optimization and continuous evolution of personality through user feedback, thereby enhancing personalized and immersive experiences.

[0052] The method provided by this invention can achieve consistent personality performance when multiple agents collaborate, and continuously optimize itself through interaction with users to meet users' personalized needs and improve user experience.

[0053] Specifically, in S2, a multi-level personality vector model is constructed based on user interaction history, semantic features, and behavioral preferences. The personality vector is constructed through the personality vector model. The personality vector does not operate independently of the task agent, but is distributed to each task agent as a global control condition, so that each task agent automatically embeds personality constraints when performing task reasoning and decision-making.

[0054] Furthermore, the multi-level personality vector model includes the following three layers:

[0055] Core personality layer: long-term stable traits, including values ​​and tone of voice;

[0056] Task-adaptive personality layer: Adjustable characteristics related to specific tasks;

[0057] Scene dynamic layer: Personality parameters that are adjusted in real time according to the environment and situation.

[0058] In S3, the personality consistency agreement includes:

[0059] Constraint format: Specifies the representation and transmission method of personality vectors among task agents, and ensures that different task agents receive the same vector with consistent meaning; for example, the representation and transmission method is standardized to a unified JSON or embedding.

[0060] The PCP (Conformity Check) definition rule is to ensure that the results generated by each task agent conform to personality constraints. For example, if the emotional temperature is ≥0.7, all responses must have a "soothing tone".

[0061] Conflict resolution: When two task agents have a style conflict while handling the same task, PCP provides a priority arbitration strategy.

[0062] Dynamic adjustment: When switching scenarios (such as from a home scenario to a car scenario), the method of shifting personality parameters is specified, while ensuring that the core personality traits remain unchanged.

[0063] In S5, user feedback includes explicit feedback and implicit feedback. Explicit feedback includes rating, modification, correction, and selection. Implicit feedback includes interaction frequency, dwell time, emotional fluctuation, and emotional response.

[0064] Furthermore, the automatic personality optimization mechanism includes:

[0065] Feedback collection: Simultaneously collect both explicit and implicit feedback from users as input signals for personality vector optimization;

[0066] Reward modeling: The reward signal is calculated by combining factors such as task success rate, human-computer interaction quality and personality consistency, and a penalty term is introduced to suppress deviations that do not conform to the personality consistency protocol;

[0067] Layered updates: Different update rates are set according to the hierarchy of personality vectors (core personality layer, task-adaptive personality layer, and scene dynamic layer) to ensure the stability of core personality traits, while allowing task and scene-related parameters to adapt quickly.

[0068] Consistency constraints: Hard and soft constraints of the personality consistency protocol are introduced during the optimization process to avoid abnormal drift of personality vectors and ensure the uniformity of output style of multiple agents;

[0069] Steady-state control: By using confidence weighting, anomaly monitoring and snapshot rollback mechanisms, the impact of single noise feedback is reduced, and the system rolls back to a steady state when it deviates from a preset threshold.

[0070] Through the above mechanism, companion AI can dynamically adjust itself according to the user's long-term habits and immediate needs while maintaining personality consistency, thereby achieving continuous optimization and personalized growth.

[0071] This invention provides a system for maintaining and automatically optimizing the personality consistency of companion AI. The system is established by employing the aforementioned method for maintaining and automatically optimizing the personality consistency of companion AI.

[0072] In summary, compared with the prior art, the present invention has the following advantages:

[0073] (1) Ensuring consistency of personality among multiple agents: Through personality vector modeling and the Personality Consistency Protocol (PCP), each task agent can share the same set of personality traits when performing different tasks, thereby avoiding inconsistencies in output tone, values, and behavioral styles. When interacting with the system, users will continuously experience a consistent personality, significantly improving the naturalness and immersion of human-computer interaction.

[0074] (2) Improve the efficiency of multi-agent collaboration: The introduction of a personality coordination and conflict arbitration module enables different task agents to quickly resolve tone and style conflicts when handling complex tasks, reducing redundant calculations and repetitive outputs. The overall operating efficiency of the system and the stability of the user experience are both improved.

[0075] (3) Support for personalization and long-term adaptation: Through the adaptive optimization mechanism of personality vector, the system can automatically adjust personality parameters based on explicit and implicit feedback during long-term use by users. This process ensures the stability of core personality traits and allows for rapid adaptation between the task layer and the scenario layer, thereby achieving personalization and continuous growth of personality.

[0076] (4) Avoid personality drift and abnormal output: ensure the long-term stability of the personality performance of the companion AI and avoid personality mutation or style abnormality.

[0077] (5) Enhanced Application Value: The method provided by this invention can be widely applied to scenarios such as intelligent assistants, virtual companions, educational robots, and medical care. The ability to maintain consistency in personality and dynamically optimize in different scenarios can significantly enhance users' trust and long-term user stickiness.

[0078] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A method for maintaining and automatically optimizing personality consistency in companion-type AI, characterized in that, Includes the following steps: Add a personal intelligent agent to companion AI; A multi-level personality vector model is constructed in the personality intelligent agent, and personality vectors are constructed through the multi-level personality vector model; the multi-level personality vector model includes a core personality layer, a task-adaptive personality layer and a scene dynamic layer; A personality consistency protocol is incorporated into the personality agent. The personality consistency protocol includes constraint format, consistency verification PCP definition rules, conflict resolution, and dynamic control. A personality coordination and conflict arbitration module is added to the personality intelligence agent. The personality coordination and conflict arbitration module compares and corrects the output results of different task intelligence agents. An automatic personality optimization mechanism is added to the personality intelligent agent. During operation, the automatic personality optimization mechanism continuously collects user feedback. The feedback is input into the personality vector optimizer through an adaptive feedback loop, driving the multi-level personality vector model to perform dynamic iterative updates.

2. The method for maintaining and automatically optimizing the personality consistency of companion AI as described in claim 1, characterized in that, A multi-level personality vector model is constructed based on user interaction history, semantic features, and behavioral preferences. Personality vectors are then constructed through the personality vector model, and these personality vectors are distributed as global control conditions to each task agent. Furthermore, each task agent automatically embeds personality constraints when performing task reasoning and decision-making.

3. The method for maintaining and automatically optimizing the personality consistency of companion AI as described in claim 2, characterized in that, The multi-level personality vector model includes the following three layers: Core personality layer: long-term stable traits, including values ​​and tone of voice; Task-adaptive personality layer: Adjustable characteristics related to specific tasks; Scene dynamic layer: Personality parameters that are adjusted in real time according to the environment and situation.

4. The method for maintaining and automatically optimizing the personality consistency of companion AI as described in claim 1, characterized in that, The Persona Consistency Protocol, or PCP for short, is a protocol for maintaining personal consistency.

5. The method for maintaining and automatically optimizing the personality consistency of companion AI as described in claim 4, characterized in that, The Personality Consistency Agreement includes: Constraint format: Specifies the way personality vectors are represented and transmitted between task agents, and ensures that different task agents receive the same vector with consistent meaning; Consistency check (PCP) definition rules: Ensure that the results generated by each task agent conform to personality constraints; Conflict resolution: When two task agents have a style conflict while handling the same task, PCP provides a priority arbitration strategy. Dynamic adjustment: When switching scenarios, the method of shifting personality parameters is specified, while ensuring that core personality traits remain unchanged.

6. The method for maintaining and automatically optimizing the personality consistency of companion AI as described in claim 5, characterized in that, The representation and transmission methods are standardized into a unified JSON or embedding.

7. The method for maintaining and automatically optimizing the personality consistency of companion AI as described in claim 1, characterized in that, User feedback includes explicit feedback and implicit feedback. Explicit feedback includes ratings, modifications, corrections, and selections. Implicit feedback includes interaction frequency, dwell time, emotional fluctuations, and emotional reactions.

8. The method for maintaining and automatically optimizing the personality consistency of companion AI as described in claim 7, characterized in that, The automatic personality optimization mechanism includes: Feedback collection: Simultaneously collect both explicit and implicit feedback from users as input signals for personality vector optimization; Reward modeling: Task success rate, human-computer interaction quality and personality consistency are combined to calculate the reward signal, and a penalty term is introduced to suppress the deviation that does not conform to the personality consistency protocol; Layered updates: Different update rates are set according to the levels of the personality vector to ensure the stability of core personality traits, while allowing task- and scenario-related parameters to adapt quickly; Consistency constraints: Hard and soft constraints of the personality consistency protocol are introduced during the optimization process to avoid abnormal drift of personality vectors and ensure the uniformity of output style of multiple agents; Steady-state control: By using confidence weighting, anomaly monitoring and snapshot rollback mechanisms, the impact of single noise feedback is reduced, and the system rolls back to a steady state when it deviates from a preset threshold.

9. A companion AI system for maintaining and automatically optimizing personality consistency, characterized in that, A system for maintaining and automatically optimizing the personality consistency of companion AI is established by adopting the personality consistency maintenance and automatic optimization method of companion AI as described in any one of claims 1-8.

Images