Intelligent agent autonomous decision interaction method based on dynamic personality and bionic memory system

Abstract

The invention provides an agent autonomous decision interaction method based on a dynamic personality and bionic memory system, and relates to the technical field of artificial intelligence and cognitive computing. Comprising the following steps: initializing multi-dimensional personality parameters of an intelligent agent, constructing a personality map, establishing a structured memory library containing scene, semantic and emotional memory, and adding a timestamp, an emotional weight and a label to the memory; environment input is received in real time, intention recognition and emotion evaluation are performed in combination with personality states and associated memories to generate a candidate behavior set, target behaviors are selected from the candidate set through a decision-making mechanism for comprehensively evaluating expected emotion benefits, personality consistency and task utility, and an interaction instruction is generated. And finally, according to behavior feedback, dynamically adjusting personality parameters and emotional states, and carrying out enhanced storage and selective forgetting on the memory bank. According to the method, personality consistency maintenance, memory-driven autonomous behavior decision and long-term personality evolution of the intelligent agent in interaction are realized, and the emotional credibility and active adaptability of personification interaction are improved.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and cognitive computing technology, and in particular to an intelligent agent autonomous decision-making and interaction method based on dynamic personality and bionic memory system. Background Technology

[0002] With the rapid development of artificial intelligence technology, interactive intelligent agents, such as virtual assistants, digital humans, non-player characters in games, and social robots, have been widely applied in various scenarios. Currently, these systems mostly rely on traditional technical approaches such as predefined rules, dialogue models trained on large amounts of data, or finite state machines to achieve interactive functions. For example, dialogue systems based on large language models can generate fluent and seemingly reasonable text responses, while characters in games typically execute preset action sequences according to scripts or behavior trees. These technologies, to a certain extent, meet basic functional requirements.

[0003] However, when people expect intelligent agents to go beyond simple command responses and exhibit more quasi-social interactions closer to human partners, the limitations of existing technologies become glaringly apparent. A core flaw lies in the fragmentation and inconsistency of personality representation. Many systems may be able to simulate certain personality traits in a single conversation, but their performance lacks stability and continuity across time and conversations. A user might experience humor from an agent in one interaction, only to receive a stereotypical response in the same situation in the next. This "amnesia" and jumps in personality state severely undermine the credibility and emotional depth of the interaction.

[0004] Furthermore, the behavioral decisions of existing intelligent agents heavily rely on external explicit instructions or clear triggering conditions, essentially remaining in a passive "stimulus-response" mode. They lack a continuous perception and understanding of the deep context and potential needs of human users, and are even less capable of proactively initiating natural social behaviors such as care, sharing, or suggestions at appropriate times. Their interactive behaviors are often fragmented and responsive, failing to form proactive interactive narratives with intrinsic motivation and coherent context. This makes the interactive experience mechanical and rigid, making it difficult to establish genuine emotional connections.

[0005] Another significant problem is the disconnect between emotional responses and the underlying personality state. While some systems integrate emotion recognition or emotion computing modules, capable of identifying user emotions and generating corresponding emotion tags or standardized responses, these emotional responses are often superficial and reflexive. They are not deeply bound to a continuously evolving core of inner personality, and emotional expression lacks personalization and historical consistency. Intelligent agents cannot, like humans, generate personalized emotional responses to events and make corresponding decisions based on their unique personality experiences, thus struggling to form a convincing sense of "self."

[0006] Furthermore, most existing systems lack the capacity for long-term memory and personality evolution. They typically treat each interaction as an independent event or only perform brief context caching. Intelligent agents cannot accumulate experience or form stable behavioral preferences from a long history of interactions, and their "personality" fails to exhibit growth, change, or adaptation consistent with living organisms after system restarts or long-term operation. This static characteristic causes stagnant interaction relationships, and user engagement rapidly declines as novelty wears off. In conclusion, current technology struggles to construct high-level intelligent agents with stable personalities, autonomous behavioral motivations, emotional rationality, and long-term growth potential, limiting their application in cutting-edge scenarios requiring deep emotional companionship, complex narrative generation, or anthropomorphic collaboration.

[0007] Furthermore, from a cognitive architecture perspective, the decision-making process of existing intelligent agents is mostly a single, linear computational process, lacking simulation of the dynamic collaboration and balance between the two modes of human thinking: "intuitive fast thinking" and "rational slow thinking." This makes it difficult for their behavior generation mechanisms to reproduce the delicate balance between rapid emotional reactions and slow, deliberate reasoning in humans, as well as the resulting realistic and nuanced behavioral tensions, thus limiting the anthropomorphic depth and psychological credibility of the intelligent agent's decision-making process. Summary of the Invention

[0008] To address the technical problems in existing technologies, such as "fragmented personality performance of intelligent agents and lack of cross-conversation consistency", "behavioral decisions rely on external instructions and lack initiative and autonomy", "emotional responses are disconnected from internal personality states, making it difficult to form a credible self", and "the system lacks long-term memory and personality evolution capabilities", this invention provides an intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory systems.

[0009] The technical solution provided by this invention is as follows: The present invention provides an intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system, comprising: S10. Personality Initialization and Modeling: Based on user settings or historical interaction data, initialize the agent's multidimensional personality parameters and construct an initial personality map; S20. Construction and updating of bionic memory system: Establish a structured memory bank, which includes at least episodic memory, semantic memory and emotional memory, and attach timestamps, emotional weights and associated tags to memory entries; S30. Autonomous Decision Triggering and Evaluation: Real-time reception of environmental input information, combined with current personality state and associated memory content, and intention recognition and emotional state evaluation through a dual-path collaborative mechanism of fast thinking and slow thinking that simulates human cognition, generating a set of candidate behaviors; the fast thinking path is based on personality map and emotional memory for rapid matching, and the slow thinking path is activated to perform deep reasoning when complexity conditions are met. S40. Behavioral Decision and Generation: Based on the preset utility function and personality consistency assessment rules, select the target behavior from the candidate behavior set and generate the corresponding interactive behavior instruction; S50. Personality Evolution and Memory Enhancement: Based on environmental feedback after the behavior is performed, personality parameters are dynamically adjusted and emotional states are updated, while memory bank is reinforced for storage and selective forgetting.

[0010] Furthermore, in step S10, the personality parameters are set based on the Big Five personality model OCEAN, including the dimensions of openness, conscientiousness, extraversion, agreeableness, and neuroticism; the personality map is constructed by mapping each personality dimension with emotional response rules and behavioral tendencies.

[0011] Furthermore, in step S20, the memory bank is stored using a graph data structure, with memory entries as nodes. Edge relationships are constructed through associated tags and sentiment weights to achieve associative memory retrieval based on semantics and sentiment.

[0012] Furthermore, in step S30, the environmental input information includes user dialogue text, environmental state variables, and predefined event triggers; the activation conditions of the slow thinking path include: the input information involves multi-task conflict, emotional ambiguity, contradictory historical feedback, or touches preset safety keywords.

[0013] Furthermore, in step S40, the utility function comprehensively evaluates the expected emotional benefits, personality consistency, and task completion utility of the candidate behavior; the personality consistency assessment is achieved by calculating the deviation of the candidate behavior from the historical behavior pattern in the personality dimension.

[0014] Furthermore, step S40 also includes introducing a controllable randomness factor into the decision-making process to simulate the uncertainty of human behavior, wherein the strength of the randomness factor is modulated by the current neurotic personality dimension parameter.

[0015] Furthermore, in step S50, the dynamic adjustment of personality parameters is based on long-term behavioral feedback trends, and a personality baseline stability mechanism is introduced to prevent personality from undergoing sudden changes under short-term feedback; the selective forgetting mechanism dynamically determines the forgetting probability based on the memory item time, emotional weight decay, and access frequency.

[0016] Furthermore, prior to step S30, a deep contextual understanding step is included: by analyzing the user's historical interaction sequence, the current session context, and related events in the memory bank, a dynamic user context model is constructed, which serves as an important input for triggering autonomous decision-making.

[0017] Furthermore, the method also includes a proactive interaction timing judgment step: by monitoring the user's idle time, historical interaction patterns and the salience of current emotional memories, the appropriateness score of proactive interaction is calculated. When the score exceeds a threshold, the process of steps S30 to S40 is automatically triggered to generate proactive interaction behavior.

[0018] Furthermore, the method also includes a safety and ethical constraint step: in the behavior decision in step S40, all candidate behaviors are screened through predefined safety rules and ethical norms filters to eliminate behavior options that do not meet the constraints, ensuring that the generated behavior instructions conform to the preset safety boundaries.

[0019] The beneficial effects of the technical solution provided by this invention include at least the following: (1) In this invention, by constructing a dynamic personality map and deeply associating it with a structured biomimetic memory system, the problem of fragmented personality expression in intelligent agents is effectively solved. The personality map is built based on multidimensional personality parameters and mapped with emotional and behavioral rules, ensuring that personality traits are computationally quantifiable and triggerable. Emotional and contextual memories in the memory bank provide historical basis for each decision. This allows the agent's emotional responses and behavioral tendencies to always revolve around its core personality parameters and maintain a high degree of coherence and consistency in different interactive sessions. Users perceive an individual with stable personality characteristics, rather than a collection of random reactions, thereby greatly enhancing the emotional credibility and anthropomorphic experience of the interaction and laying the foundation for establishing long-term relationships.

[0020] (2) In this invention, by integrating deep contextual understanding of the environment, memory-driven decision-making, and judgment of the timing of proactive interaction, the intelligent agent is endowed with true behavioral autonomy. The system not only passively responds to user input but also continuously analyzes the user's state, historical patterns, and emotional memories, and proactively initiates the interaction process when it is deemed appropriate. The decision-making process combines current personality tendencies with emotional associations from past memories, making proactively initiated behaviors (such as care, sharing, or suggestions) inherently logical and emotionally motivated, rather than mechanically triggered at set times. This breaks the traditional passive "stimulus-response" model, enabling the intelligent agent to engage in natural, timely, and meaningful proactive communication like a human partner, significantly enhancing the vividness of the interaction and the user's immersion.

[0021] (3) In this invention, by designing a personality evolution and memory enhancement mechanism, the intelligent agent is endowed with the ability to learn and grow over a long period of time. The personality parameters are not fixed, but are smoothly adjusted according to long-term behavioral feedback trends and protected by a stability mechanism to avoid sudden changes. The bionic memory system enhances important experiences through emotional weights and selectively forgets based on time, decay rate, and access frequency, simulating the characteristics of human memory. This dynamic update mechanism allows the intelligent agent's personality preferences, behavioral patterns, and understanding of users to continuously evolve and deepen with the interaction process. Users can feel the "growth" and "change" of the intelligent agent, thereby establishing a deeper and more sticky interactive relationship and expanding the application value of the system in complex scenarios such as long-term companionship and personalized adaptation.

[0022] (4) In this invention, a biological rationality simulation of complex decision-making processes is achieved by introducing a "fast thinking-slow thinking" dual-path collaborative decision-making mechanism. The fast thinking path ensures efficient and anthropomorphic immediate response to familiar situations, while the slow thinking path provides prudent reasoning ability for complex and high-risk decisions. The synergy between the two paths (such as the inhibition and correction of fast thinking impulses by slow thinking) makes the decision-making process of the intelligent agent exhibit the hesitation, weighing and self-control unique to humans, greatly enhancing the psychological realism and logical consistency of the behavioral output, and solving the technical problem of the mechanization of behavior in a single decision-making process. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the autonomous decision-making interaction method for intelligent agents based on dynamic personality and bionic memory systems, provided in an embodiment of the present invention. Detailed Implementation

[0025] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0026] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0027] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0028] In embodiments of the present invention, sometimes the subscript is as follows: It may be mistakenly written as a non-subscript form such as W1. When the distinction is not emphasized, the meaning they express is the same.

[0029] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0030] Reference manual attached Figure 1 The diagram illustrates a flowchart of an intelligent agent autonomous decision-making interaction method based on a dynamic personality and bionic memory system, provided in an embodiment of the present invention.

[0031] This invention provides an intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system. The processing flow may include the following steps: S10. Personality Initialization and Modeling: Based on user settings or historical interaction data, initialize the agent's multidimensional personality parameters and construct an initial personality map.

[0032] To construct an intelligent agent with self-awareness and autonomous decision-making capabilities, the core of the approach lies in a localized, continuously running context-aware decision-making loop. This method begins with personality initialization and modeling. Based on explicit user-provided settings or by analyzing historical interaction data, the system initializes a series of intrinsic personality dimension parameters for the intelligent agent. Based on these parameters, the system constructs an initial personality map, which defines the fundamental mapping relationship between personality traits and subsequent emotional response rules and behavioral tendencies, establishing the initial tone for the agent's behavior and reactions.

[0033] S20. Construction and updating of bionic memory system: Establish a structured memory bank, which includes at least episodic memory, semantic memory and emotional memory, and attach timestamps, emotional weights and related tags to memory entries.

[0034] After establishing the foundation of personality, the system constructs and updates a bionic memory system. The system establishes a structured memory bank that distinguishes and stores at least three types of memories: episodic memories recording specific events and experiences, semantic memories storing facts and concepts, and emotional memories associated with emotional experiences. Each stored memory entry is attached with a precise timestamp, an emotional weight indicating emotional intensity and whether it is positive or negative, and tags used to describe and associate the content, thus forming a continuously expandable and retrieval-enabled memory system.

[0035] S30. Autonomous Decision Triggering and Evaluation: Real-time reception of environmental input information, combined with current personality state and associated memory content, and intention recognition and emotional state evaluation through a dual-path collaborative mechanism simulating human cognition of fast thinking and slow thinking, generating a set of candidate behaviors; the fast thinking path is based on personality map and emotional memory for rapid matching, and the slow thinking path is activated to perform deep reasoning when complexity conditions are met.

[0036] The core step S30 of this invention, autonomous decision-making triggering and evaluation, is achieved through a "dual-path collaborative processing mechanism" that simulates human cognition. This mechanism processes environmental inputs in parallel: Fast Thinking Path: Input information is first processed through the Fast Thinking Path. This path is directly coupled with personality mapping and affective memory mapping, generating preliminary behavioral intentions based on "intuition" and "personality habits" in a very short time through efficient parallel pattern matching. For example, when a highly agreeable personality senses a user's frustration, the Fast Thinking Path will instantly map out the intention to "express comfort".

[0037] Slow Thinking Triggering and Processing: Simultaneously, the system assesses the complexity of the current decision. If the situation is judged to be simple and low-risk (such as a routine greeting), the intention of the fast thinking path can be directly output after minor adjustments. If complexity triggering conditions are met (such as detecting contradictory user instructions, dialogue involving ethical boundaries, or poor feedback from similar historical situations), the slow thinking path is activated. The slow thinking path invokes complete semantic memory and a deep contextual model, focusing attention on sequential reasoning. One of its key functions is to "secondarily review" the initial intention of the fast thinking path: it may adopt and enhance its rationality, modify its expression and intensity, or suppress it if it is found to seriously conflict with safety rules or long-term goals, generating entirely new and more prudent behavioral options.

[0038] Collaborative Output and Personality Regulation: Ultimately, one or more behavioral intentions processed through dual paths (or determined to require only fast path processing) converge to form a candidate behavior set, which is then delivered to subsequent decision-making steps. Crucially, the activation sensitivity of the slow thinking path and its intervention in fast thinking intentions are not fixed values, but are dynamically adjusted by the agent's current personality dimensions of "neuroticism" (influencing sensitivity to uncertainty and negative feedback) and "conscientiousness" (influencing the depth of consideration for rules and consequences). This transforms personality traits from static parameters into a dynamic engine that directly influences cognitive processing style.

[0039] S40. Behavioral Decision-Making and Generation: Based on the preset utility function and personality consistency assessment rules, select the target behavior from the candidate behavior set and generate the corresponding interactive behavior instructions.

[0040] Next, the system proceeds to the behavior decision-making and generation step. The system uses a pre-defined utility function to evaluate each behavior in the candidate behavior set. This utility function calculates the overall utility that each behavior may bring. Simultaneously, based on personality consistency assessment rules, the system measures the degree to which each candidate behavior conforms to the agent's established personality traits. By comprehensively comparing these evaluation results, the system selects the optimal target behavior from the candidate set and transforms it into a specific, executable interactive behavior instruction. This instruction can be a dialogue response, emotional expression, proactive service, or task execution, etc.

[0041] S50. Personality Evolution and Memory Enhancement: Based on environmental feedback after the behavior is performed, personality parameters are dynamically adjusted and emotional states are updated, while memory bank is reinforced for storage and selective forgetting.

[0042] Finally, the system executes personality evolution and memory reinforcement steps. After the agent performs a behavior and receives environmental feedback, the system dynamically adjusts the weights of its personality dimension parameters and updates its current emotional state based on the feedback. Simultaneously, the bionic memory system reinforces and stores relevant memory entries based on the emotional intensity and importance of the interaction, deepening memory; and selectively forgets outdated or low-weight memories according to a certain mechanism. This process allows the agent's personality traits and memory content to continuously evolve over time and with deeper interaction, forming a unique developmental trajectory.

[0043] In one possible implementation, in step S10, the personality parameters are set based on the Big Five personality model (OCEAN), including the dimensions of openness, conscientiousness, extraversion, agreeableness, and neuroticism; the personality map is constructed by mapping each personality dimension with emotional response rules and behavioral tendencies.

[0044] Regarding the setting of personality parameters, in one implementation, the system uses the Big Five personality model as the theoretical basis for personality modeling. The initialized multidimensional personality parameters specifically include five core dimensions: openness, conscientiousness, extraversion, agreeableness, and neuroticism. The system assigns an initial weight value to each dimension. The constructed personality map is essentially a mapping network that associates and maps each of the aforementioned personality dimensions with a series of specific emotional response rules and behavioral tendencies, thereby transforming abstract personality traits into calculable and triggerable emotional and behavioral rules.

[0045] In one possible implementation, in step S20, the memory bank is stored using a graph data structure, with memory entries as nodes. Edge relationships are constructed through associated tags and sentiment weights to achieve associative memory retrieval based on semantics and sentiment.

[0046] Regarding the structure of the biomimetic memory system, in one implementation, the system employs a graph data structure to realize a structured memory bank. Each independent memory entry is stored as a node in the graph. The edges between nodes are constructed using association labels and sentiment weights, where association labels describe the semantic or logical connections between memory contents, and sentiment weights characterize the strength of the emotional association between memories. This graph structure can efficiently support associative memory retrieval based on semantic similarity and emotional resonance, enabling the system to start from a single memory and find a series of related memories.

[0047] In one possible implementation, in step S30, the environmental input information includes user dialogue text, environmental state variables, and predefined event triggers; the activation conditions for the slow thinking path include: the input information involves multi-task conflict, emotional ambiguity, contradictory historical feedback, or triggering preset safety keywords.

[0048] Regarding the environmental input and evaluation mechanism, the environmental input information received by the system is diverse, mainly including dialogue text input by the user through natural language, environmental state variables provided by sensors or system status, and predefined event triggers triggered by internal logic or external events. When performing intent recognition and emotional state evaluation, the system uses a dual-path collaborative mechanism: the fast thinking path quickly calls upon the personality map and emotional memory for initial evaluation; if the slow thinking path is activated, it further integrates personality tendency weights with broader semantic memory and situational models to generate evaluation conclusions that better reflect the agent's "personality" and "experiences."

[0049] In one possible implementation, in step S30, after the slow thinking path is activated, the initial intention generated by the fast thinking path is verified, suppressed, or deepened, and the intensity of the intervention is modulated by the neuroticism and conscientiousness dimensions in the current personality parameters.

[0050] Regarding the regulatory mechanism between the two brain pathways, the slow thinking pathway's intervention in the fast thinking pathway is neither entirely nor entirely absent. The intensity of this intervention is a continuous variable, jointly determined by the agent's current "neuroticism" and "conscientiousness" personality dimensions. Specifically, a higher "neuroticism" score lowers the activation threshold of slow thinking, making it more susceptible to in-depth scrutiny of uncertainty and potential negative feedback; while a higher "conscientiousness" score strengthens the intervention of slow thinking, enabling it to more rigorously suppress or modify the impulsive intentions of fast thinking based on rules and long-term goals. This dynamic regulatory mechanism allows personality traits to be directly internalized into the agent's cognitive processing style.

[0051] In one possible implementation, in step S40, the utility function comprehensively evaluates the expected emotional benefits, personality consistency, and task completion utility of the candidate behavior; the personality consistency assessment is achieved by calculating the deviation of the candidate behavior from the historical behavior pattern in terms of personality dimensions.

[0052] Specifically, regarding the evaluation rules for behavioral decisions, the system's preset utility function is a multi-objective evaluation model. It simultaneously calculates the expected emotional benefits of a candidate behavior, the degree of consistency between the behavior and the agent's current personality, and the task completion utility directly related to the behavior. The assessment of personality consistency is achieved through a specific calculation process: calculating the deviation between the candidate behavior's performance across various personality dimensions and the baseline formed by the agent's historical behavioral patterns in those dimensions; the smaller the deviation, the higher the consistency.

[0053] To simulate the inherent uncertainty in human behavior, a controllable randomness factor is introduced into the system's decision-making process. This factor introduces a small perturbation to the evaluation result during the final decision, creating a degree of behavioral unpredictability. The strength of this randomness factor is not fixed but dynamically adjusted by the agent's current neuroticism dimension parameters; the higher the neuroticism level, the greater the randomness of the decision may be.

[0054] In one possible implementation, in step S50, the dynamic adjustment of personality parameters is based on long-term behavioral feedback trends, and a personality baseline stability mechanism is introduced to prevent personality from undergoing sudden changes under short-term feedback; the selective forgetting mechanism dynamically determines the forgetting probability based on the memory item time, emotional weight decay, and access frequency.

[0055] Regarding the detailed mechanisms of personality evolution and memory management, the dynamic adjustment of personality parameters does not simply depend on a single feedback, but rather on a smooth adjustment based on the trends presented by long-term behavioral feedback. The system introduces a personality baseline stability mechanism, which acts as a buffer to effectively prevent unreasonable mutations in personality parameters due to short-term, drastic feedback. In terms of memory management, the selective forgetting mechanism dynamically determines the forgetting probability of each memory item based on three core elements: the length of time the memory has been stored, the degree to which emotional weight naturally decays over time, and the frequency with which the memory is accessed.

[0056] As an optional implementation, before step S30, a deep contextual understanding step is also included: by analyzing the user's historical interaction sequence, the current session context, and related events in the memory bank, a dynamic user context model is constructed, which serves as an important input for triggering autonomous decision-making.

[0057] Before the core decision-making process is triggered, the system includes a preparatory step of deep contextual understanding. This step constructs a dynamic, multi-dimensional user context model by comprehensively analyzing the user's long-term historical interaction sequences, the contextual information of the current session, and related events retrieved from the memory bank. This deep contextual model is then used as one of the key inputs into the autonomous decision-making triggering and evaluation steps, enabling decisions to be based on richer background cognition.

[0058] In one possible implementation, the method further includes a proactive interaction timing judgment step: by monitoring the user's idle time, historical interaction patterns and the salience of current emotional memories, a suitability score for proactive interaction is calculated. When the score exceeds a threshold, the process of steps S30 to S40 is automatically triggered to generate proactive interaction behavior.

[0059] To achieve truly proactive interaction, the system also independently runs a proactive interaction timing judgment step. This step continuously monitors the user's idle time, analyzes the temporal patterns and regularities of the user's historical interactions, and calculates the activation level of highly salient events in the current emotional memory. Combining these factors, the system calculates an appropriateness score for proactive interaction. When this score exceeds a preset threshold, the system will autonomously trigger the subsequent complete decision-making process from contextual awareness to behavior generation, thus initiating a natural interaction without external instructions.

[0060] In one possible implementation, the method further includes a safety and ethical constraint step: in the behavior decision in step S40, all candidate behaviors are screened through predefined safety rules and ethical norms filters to eliminate behavior options that do not meet the constraints, ensuring that the generated behavior instructions conform to the preset safety boundaries.

[0061] To ensure the safety and reliability of the agent's behavior, the system integrates a safety and ethical constraint step into the behavior decision-making process. All candidate behaviors generated through utility function evaluation must be screened through a predefined safety rule and ethical norm filter before being ultimately selected. This filter rigorously examines the potential impact of behavioral instructions and forcibly eliminates any behavioral options that do not comply with the established safety and ethical constraints, thereby ensuring that the final generated and executed behavioral instructions always remain within the preset safety boundaries.

[0062] In addition to the above-described embodiments, the technical solution of this invention can also be implemented in other alternative ways. Regarding personality modeling, besides the Big Five personality model, other psychological personality models can be used as a foundation, such as the MBTI 16 personality index or the Enneagram 9 personality model, to initialize personality dimensions and construct a personality map. In the implementation of the bionic memory system, a dedicated graph database can be used to store and manage memory entries and their relationship networks to further optimize memory association, retrieval, and associative efficiency. In the specific implementation of the dual-path collaborative mechanism, the fast thinking path can use lightweight neural networks or rule engines to achieve rapid response, while the slow thinking path can integrate symbolic reasoning, reinforcement learning algorithms, or call large-scale pre-trained models for deep causal analysis, forming a hybrid intelligent decision-making paradigm. At the decision-making mechanism level, reinforcement learning algorithms can be introduced, enabling the behavioral decision-making model to self-optimize and adjust through reward signals in long-term interaction with the environment, thereby improving the long-term adaptability and effectiveness of decisions. In the behavior generation stage, advanced generative pre-trained models can be combined to generate richer and more natural language behavior descriptions, which are then parsed and converted into specific instructions that the system can execute.

[0063] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: (1) In this invention, by constructing a dynamic personality map and deeply associating it with a structured biomimetic memory system, the problem of fragmented personality expression in intelligent agents is effectively solved. The personality map is built based on multidimensional personality parameters and mapped with emotional and behavioral rules, ensuring that personality traits are computationally quantifiable and triggerable. Emotional and contextual memories in the memory bank provide historical basis for each decision. This allows the agent's emotional responses and behavioral tendencies to always revolve around its core personality parameters and maintain a high degree of coherence and consistency in different interactive sessions. Users perceive an individual with stable personality characteristics, rather than a collection of random reactions, thereby greatly enhancing the emotional credibility and anthropomorphic experience of the interaction and laying the foundation for establishing long-term relationships.

[0064] (2) In this invention, by integrating deep contextual understanding of the environment, memory-driven decision-making, and judgment of the timing of proactive interaction, the intelligent agent is endowed with true behavioral autonomy. The system not only passively responds to user input but also continuously analyzes the user's state, historical patterns, and emotional memories, and proactively initiates the interaction process when it is deemed appropriate. The decision-making process combines current personality tendencies with emotional associations from past memories, making proactively initiated behaviors (such as care, sharing, or suggestions) inherently logical and emotionally motivated, rather than mechanically triggered at set times. This breaks the traditional passive "stimulus-response" model, enabling the intelligent agent to engage in natural, timely, and meaningful proactive communication like a human partner, significantly enhancing the vividness of the interaction and the user's immersion.

[0065] (3) In this invention, by designing a personality evolution and memory enhancement mechanism, the intelligent agent is endowed with the ability to learn and grow over a long period of time. The personality parameters are not fixed, but are smoothly adjusted according to long-term behavioral feedback trends and protected by a stability mechanism to avoid sudden changes. The bionic memory system enhances important experiences through emotional weights and selectively forgets based on time, decay rate, and access frequency, simulating the characteristics of human memory. This dynamic update mechanism allows the intelligent agent's personality preferences, behavioral patterns, and understanding of users to continuously evolve and deepen with the interaction process. Users can feel the "growth" and "change" of the intelligent agent, thereby establishing a deeper and more sticky interactive relationship and expanding the application value of the system in complex scenarios such as long-term companionship and personalized adaptation.

[0066] (4) In this invention, a biological rationality simulation of complex decision-making processes is achieved by introducing a "fast thinking-slow thinking" dual-path collaborative decision-making mechanism. The fast thinking path ensures efficient and anthropomorphic immediate response to familiar situations, while the slow thinking path provides prudent reasoning ability for complex and high-risk decisions. The synergy between the two paths (such as the inhibition and correction of fast thinking impulses by slow thinking) makes the decision-making process of the intelligent agent exhibit the hesitation, weighing and self-control unique to humans, greatly enhancing the psychological realism and logical consistency of the behavioral output, and solving the technical problem of the mechanization of behavior in a single decision-making process.

[0067] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0068] The following points need to be explained: (1) The accompanying drawings of the embodiments of the present invention only involve the structures involved in the embodiments of the present invention. Other structures can refer to the general design.

[0069] (2) For clarity, the thickness of layers or regions is enlarged or reduced in the drawings used to describe embodiments of the invention, i.e., these drawings are not drawn to scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being “above” or “below” another element, the element may be “directly” located “above” or “below” the other element or there may be intermediate elements.

[0070] (3) Where there is no conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0071] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system, characterized in that, include: S10. Personality Initialization and Modeling: Based on user settings or historical interaction data, initialize the agent's multidimensional personality parameters and construct an initial personality map; S20. Construction and updating of bionic memory system: Establish a structured memory bank, which includes at least episodic memory, semantic memory and emotional memory, and attach timestamps, emotional weights and associated tags to memory entries; S30. Autonomous Decision Triggering and Evaluation: Real-time reception of environmental input information, combined with current personality state and associated memory content, and intention recognition and emotional state evaluation through a dual-path collaborative mechanism of fast thinking and slow thinking that simulates human cognition, generating a set of candidate behaviors; the fast thinking path is based on personality map and emotional memory for rapid matching, and the slow thinking path is activated to perform deep reasoning when complexity conditions are met. S40. Behavioral Decision and Generation: Based on the preset utility function and personality consistency assessment rules, select the target behavior from the candidate behavior set and generate the corresponding interactive behavior instruction; S50. Personality Evolution and Memory Enhancement: Based on environmental feedback after the behavior is performed, personality parameters are dynamically adjusted and emotional states are updated, while memory bank is reinforced for storage and selective forgetting.

2. The intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system according to claim 1, characterized in that, In step S10, the personality parameters are set based on the Big Five personality model OCEAN, including the dimensions of openness, conscientiousness, extraversion, agreeableness, and neuroticism; the personality map is constructed by mapping each personality dimension with emotional response rules and behavioral tendencies.

3. The intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system according to claim 1, characterized in that, In step S20, the memory bank is stored using a graph data structure, with memory entries as nodes. Edge relationships are constructed through associated tags and sentiment weights to achieve associative memory retrieval based on semantics and sentiment.

4. The intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system according to claim 1, characterized in that, In step S30, the environmental input information includes user dialogue text, environmental state variables, and predefined event triggers; the activation conditions of the slow thinking path include: the input information involves multi-task conflict, emotional ambiguity, contradictory historical feedback, or touches preset safety keywords.

5. The intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system according to claim 1 or 4, characterized in that, In step S30, after the slow thinking path is activated, the initial intention generated by the fast thinking path is verified, suppressed, or deepened. The intensity of the intervention is regulated by the neuroticism and conscientiousness dimensions in the current personality parameters.

6. The intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system according to claim 1 or 4, characterized in that, In step S40, the utility function comprehensively evaluates the expected emotional benefits, personality consistency, and task completion utility of the candidate behavior; the personality consistency assessment is achieved by calculating the deviation of the candidate behavior from the historical behavior pattern in the personality dimension.

7. The intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system according to claim 6, characterized in that, Step S40 also includes introducing a controllable randomness factor into the decision-making process to simulate the uncertainty of human behavior, wherein the strength of the randomness factor is modulated by the current neurotic personality dimension parameter.

8. The intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system according to claim 1, characterized in that, In step S50, the dynamic adjustment of personality parameters is based on long-term behavioral feedback trends, and a personality baseline stability mechanism is introduced to prevent personality from undergoing sudden changes under short-term feedback. The selective forgetting mechanism dynamically determines the forgetting probability based on the memory item's time, emotional weight decay, and access frequency.

9. The intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system according to claim 1, characterized in that, The method also includes a proactive interaction timing judgment step: by monitoring the user's idle time, historical interaction patterns and the salience of current emotional memories, a suitability score for proactive interaction is calculated. When the score exceeds a threshold, the process of steps S30 to S40 is automatically triggered to generate proactive interaction behavior.

10. The intelligent agent autonomous decision-making interaction method based on dynamic personality and bionic memory system according to claim 1, characterized in that, The method also includes a safety and ethical constraint step: in the behavior decision in step S40, all candidate behaviors are screened through predefined safety rules and ethical norms filters to eliminate behavior options that do not meet the constraints, ensuring that the generated behavior instructions meet the preset safety boundaries.

Images