Top Read Articles

Published in last 1 year |

In last 2 years |

In last 3 years |

All

Please wait a minute...


For Selected: Download Citations EndNote Ris BibTeX Toggle Thumbnails

Select

When artificial intelligence faces human emotions: The impact mechanism of emotion expression in AI-empowered service robots on user experience LUO Lijuan, WANG Kang, HU Jinmiao, XU Sihua Advances in Psychological Science    2025, 33 (6): 1006-1026.   DOI: 10.3724/SP.J.1042.2025.1006 Abstract （3915）   HTML （184）    PDF （1631KB）（8667）       Knowledge map    The rapid advancement of next-generation AI technologies has fundamentally reshaped interaction patterns between users and service providers. Nowadays, users not only expect AI to effectively solve problems but also aspire to gain positive emotional experiences during the interaction process. However, current AI services still face challenges such as user resistance, low acceptance, and poor service experiences. Addressing how service robots can establish effective emotional communication with users to deliver personalized, intelligent, and empathetic service experiences has become a critical research frontier. This study investigates the holistic process of user-service robot interaction through the lens of AI-empowered emotion connection, establishing an integrated framework of "Emotion recognition, Emotion understanding, Emotional connection." We propose the following research framework and systematically investigates three principal research dimensions: (1) User emotion recognition and emotion matrix construction based on a multidimensional emotion computing model. A user emotion recognition computing model is developed using machine learning algorithms and decision-level weighted fusion to resolve inconsistencies in cross-dimensional emotional expressions. Building upon the established multidimensional emotion recognition model, the valence-arousal-dominance (VAD) model is adopted as the analytical framework. Through combined machine learning and qualitative analysis methods, we systematically characterize users' emotional responses across different service stages and contexts. This research concept helps build a bridge between emotion recognition and service interaction, laying the foundation for real-time emotional responses with service interactions. (2) The impact mechanisms of AI-empowered emotional expression content on user experience from the perspective of service journey. Human-robot interaction processes can be categorized into three sequential stages: initial service encounter, service usage, and service feedback. The initial encounter stage prioritizes AI emotional expression to stimulate user interest and establish trust, while the usage stage focuses on delivering affective experiences to enhance satisfaction. The feedback stage aims to mitigate user dissatisfaction and attain forgiveness. Aligning with stage-specific objectives, we propose differentiated emotional expression strategies. Drawing on Trust Theory, Cognitive Appraisal Theory, and Basic Psychological Needs Theory, we hypothesize that service robots' implementation of stage-specific emotional expressions (positive emotion in initial encounters, empathy during service usage, and gratitude in service feedback) can systematically enhance user experience. This study delves into the underlying mechanisms of AI-empowered emotional expression content on user experience at each service stage. Moreover, we also propose three moderating factors—the anthropomorphic features of AI, time pressure, and the types of explanatory information provided—as boundary conditions in different stages. This hypothesis framework enables the systematic investigation of when and why differentiated emotional content across service journey stages impacts user experience. This research concept fosters a holistic and dynamic understanding of service journey stages, highlighting the significance of leveraging AI emotional intelligence to activate user experience throughout the journey. (3) The impact mechanisms of AI-empowered emotional expression modalities on user experience from the perspective of service contexts. Service contexts are classified into hedonic-oriented and utilitarian-oriented scenarios, where user preferences diverge significantly. Hedonic contexts center on experiential values like enjoyment, pleasure, and emotional engagement, while utilitarian contexts emphasize functional benefits including practicality, efficiency, and utility. Through the theoretical lenses of Social Presence Theory, Psychological Distance Theory, and Emotions-as-Social-Information Theory, we hypothesize that service robots' implementation of embodied emotional expression modalities (mono-sensory vs. multisensory) in hedonic-oriented and utilitarian-oriented service contexts can significantly enhance user service experience. This study further examines the underlying mechanisms of AI-empowered embodied emotional expression modalities on user experience at each service context. Moreover, we also propose two moderating factors—relationship norm orientation and task complexity—as boundary conditions in different contexts. This hypothesis framework enables the systematic investigation of when and why emotional expression modalities across distinct service contexts impact user experience. This research concept fosters differentiated thinking on the modalities of AI's emotional expression in service contexts, shedding light on the importance of emotional modalities in both hedonic-oriented and utilitarian-oriented service contexts. This study advances the understanding of emotional expression mechanisms in service robots and user experience enhancement strategies within intelligent services. It offers significant theoretical contributions and practical insights. In terms of theoretical significance, this research enriches human-AI interaction theory by proposing a comprehensive framework for service robots' emotional expression mechanisms. It empirically demonstrates how AI-driven affective expressions activate and influence user experience while clarifying underlying mechanisms, thereby advancing the theoretical foundation for emotionally intelligent interaction design. In terms of practical significance, this research provides a new direction for the integrated development of AI and service industry, enabling service providers to optimize touchpoints across the service journey. More importantly, it underscores the value of affective intelligence, providing robust support for the high-quality and sustainable development of the service robotics industry. Table and Figures | Reference | Related Articles | Metrics

Select

The bidirectional trust in the context of new human-machine relationships XIE Yubin, ZHOU Ronggang Advances in Psychological Science    2025, 33 (6): 916-932.   DOI: 10.3724/SP.J.1042.2025.0916 Abstract （2430）   HTML （128）    PDF （679KB）（5759）       Knowledge map    In the context of the rapid development of artificial intelligence, the relationship between humans and machines is shifting from the traditional “auxiliary-subordinate” model to a “equal collaboration” or even “symbiotic coevolution” model. Most of research has mainly focused on one-directional trust from humans to machines, but as intelligent agent gains greater autonomy and decision-making capabilities, mutual trust is becoming a central issue in human-machine collaboration. This paper examines the mechanisms for building mutual trust between humans and machines. It also explores measurement methods, and the practical challenges involved. The goal is to provide theoretical support for the design and optimization of future intelligent systems. Building on existing human-machine trust frameworks, this paper proposes a dynamic model of mutual trust. It introduces a three-stage model: “dispositional trust-perceived trust-behavioral trust,” which covers both human-to-machine and machine-to-human trust. The model emphasizes perceived trust as the key bridge between dispositional trust and behavioral trust, highlighting its role in the transfer of trust between AI, intelligent agents, and humans. Dispositional trust: The initial stage of trust, which is based on an individual's inherent traits and is independent of specific contexts. It lays the foundation for the subsequent development of trust. Perceived trust: Gradually formed during interaction, this stage reflects the dynamic perception of the other party's behavior, attitude, and trustworthiness. It is the core of emotional trust transfer and dynamic adjustment. Behavioral trust: The final manifestation of trust, expressed through concrete behaviors such as dependence, cooperation, and action. It is post-action trust based on behavioral feedback, reflecting the ultimate outcome of the trust relationship. The advantages of this model are reflected in several key aspects. First, its dynamic evolutionary characteristics allow the model to fully capture the development of trust from dispositional trust to perceived trust and finally to behavioral trust, effectively accommodating the complexity and variability of trust relationships in human-machine interactions. Second, the model emphasizes bidirectional trust transfer, focusing on the interaction between humans and intelligent agents. It highlights the role of perceived trust as the crucial bridge between dispositional trust and behavioral trust, offering in-depth insights into its significance in emotional trust transfer and dynamic adjustment, thus providing unique guidance for optimizing human-machine interaction. Third, the model introduces an expanded perspective on dispositional trust by incorporating algorithmic trust, exploring the sources of initial trust in algorithms and the impact of individual algorithm aversion, thereby offering a new theoretical foundation for algorithmic trust research. Lastly, the model provides an in-depth analysis of behavioral trust, emphasizing the impact of machine behavior on human-machine trust, such as the negative effect on “perceived trustworthiness” when a machine denies a human request, and revealing the emotional and behavioral consequences of trust misalignment. The purpose of this theoretical model is to develop methods for measuring and modeling human-machine mutual trust based on the characteristics of different scenarios. Building on a review of existing measurement methods and drawing from interpersonal trust measurement experience, this paper introduces a framework and methods for mutual trust between humans and machines. The study focuses on several key areas: developing stage-specific measurement tools for dispositional trust, perceived trust, and behavioral trust; exploring multidimensional, multilevel methods that combine subjective reports, physiological signals, and behavioral data to create a dynamic monitoring and calibration system; and adapting interpersonal trust quantification methods to design trust modeling tools suited to human-machine interactions. Ultimately, this research aims to provide a systematic, operable framework for measuring and modeling mutual trust, laying the foundation for dynamic evaluation and intelligent adjustment. In terms of application, this paper examines the practical value of mutual trust through case studies in autonomous driving and aviation. It also discusses current challenges, such as individual differences that hinder trust development, the lack of standardized tools for measuring machine trust, and the unclear long-term psychological effects of mutual trust on users. The paper calls for further research to refine trust measurement tools, address issues of “over-trust” or “mistrust” in human-machine trust alignment and define the boundaries of machine trust behavior within ethical and legal frameworks. By integrating theoretical and methodological innovations, this paper offers new directions for research on trust mechanisms in human-machine collaboration and provides valuable guidance for the development of efficient and safe intelligent systems. Table and Figures | Reference | Related Articles | Metrics

Select

Promoting or inhibiting? The double-edged sword effect of acceptance of generative AI advice on creativity ZONG Shuwei, YANG Fu, LONG Lirong, HAN Yi Advances in Psychological Science    2025, 33 (6): 905-915.   DOI: 10.3724/SP.J.1042.2025.0905 Abstract （2138）   HTML （105）    PDF （576KB）（4453）       Knowledge map    Generative artificial intelligence (GAI), as an emerging artificial intelligence technology, has showcased remarkable creativity across diverse domains. However, members of an organization are the main carriers of corporate innovation and creativity. A pivotal question persists in academic circles: Will the integration of GAI technology into the workplace and the promotion of its work suggestions or solutions among organizational members stimulate or stifle their creativity? This question remains largely unexplored, particularly regarding the phenomenon of advice taking from GAI within the realm of organizational management. This study aims to address this significant gap. Initially, grounded in the context of organizational management, we elucidate the connotation and dimensions of advice taking from GAI, addressing the fundamental inquiry: “what constitutes advice taking from GAI?”. Subsequently, we investigate the double-edged sword effects of advice taking from GAI on creativity, examining its effects at both the employee and team levels through the lenses of social cognition and goal orientation. We endeavor to clarify the issue, “what is the relationship between advice taking from GAI and creativity?”. Finally, we synthesize an exploration of the boundary conditions for the effect of advice taking from GAI, seeking to answer: “under what circumstances does advice taking from GAI yield positive outcomes?” This study makes three contributions. First, we explore the dimensions and measurement of the construct of advice taking from GAI, paving the way for new research directions in advice taking. In the digital era, AI has become a novel source for organizations to acquire unique information and generate personalized recommendations. Conducting in-depth academic research in this area contributes to a better understanding of the effectiveness of advice taking from GAI. Unfortunately, existing studies primarily focus on the immediate effects of GAI on idea generation, neglecting that the application of GAI in organizations is a multi-stage process, rather than a simple tool usage. In response, study 1 introduces a refined framework comprising three stages: advice solicitation, advice evaluation, and advice adoption, highlighting the iterative interaction process between organizational members and GAI. By investigating how organizational members gradually adopt GAI advice and examining the impact of this process on employee creativity, this study broadens the scope of research on the relationship between AI techniques and human creativity. It also underscores the importance of focusing on the iterative processes of screening, evaluating, adjusting, and applying AI advice and the subsequent outcomes. Second, from the approach of social cognitive and goal orientation, this study investigates the double-edged sword effect of individual/team advice taking from GAI on individual/team creativity and its underlying mechanisms, expanding the research perspective on the advice taking from GAI. A review of the literature reveals that most existing studies on AI adoption focus on its influencing factors and formation mechanisms, with limited discussion on the effects of AI adoption. In the few studies related to GAI and organizational creativity, the primary focus is on the individual-level use of GAI and its related outcomes, without systematically, comprehensively, or multi-levelly exploring the mechanisms through which advice taking from GAI influences creativity. To address this gap, study 2 and study 3, respectively, examine how the differentiated impacts of advice taking from GAI on individuals' cognitive states and teams' goal orientations drive divergent outcomes for organizational creativity. Specifically, the double-edged sword effect of advice taking from GAI on creativity is explored. This reveals the complex effect mechanism of GAI in different contexts, particularly the dynamic balance between individuals' sense of efficacy and dependence on AI, as well as the competing orientations of teams toward learning goals and performance goals. Third, this study systematically identifies corresponding intervention strategies from multiple organizational levels to enhance the positive effects of advice taking from GAI on creativity while mitigating its negative effects. Existing research on the boundary conditions of AI usage effects primarily focuses on individual-level characteristics, with limited attention to the intervention effects of organizational-level factors. As both the practical bearer of AI application outcomes in the workplace and a critical component of AI governance frameworks, organizations hold significant governance responsibilities. In this context, this study proposes intervention strategies from two perspectives: resource support and feedback mechanisms, to examine the boundary conditions under which advice taking from GAI exerts positive or negative effects on creativity. These findings help clarify the organizational levels at which interventions are needed to influence members' advice taking from GAI, providing decision-makers with valuable references. Table and Figures | Reference | Related Articles | Metrics

Select

Workplace artificial intelligence role classification: Impacts on employee psychology and behavior and coping strategies TAN Meili, YIN Xiangzhou, ZHANG Guanglei, XIONG Puzhen Advances in Psychological Science    2025, 33 (6): 933-947.   DOI: 10.3724/SP.J.1042.2025.0933 Abstract （1848）   HTML （89）    PDF （753KB）（4631）       Knowledge map    Artificial intelligence (AI) has become an integral part of modern workplaces. However, understanding its impact on employees remains a complex and dynamic area of research. Through a comprehensive review of relevant literature and taking into account the practical applications of AI, this study innovatively approaches the topic. It classifies AI's roles in the workplace from the employees' perspective into four distinct types: “opponent”, “assistant”, “colleague”, and “leader”. This sets our study apart from previous research, as we place greater emphasis on the employee-centered perspective. Based on a review of the literature, this classification is based on the “substitution-aid-augmentation-management” framework. From this perspective, the study systematically examines how these different AI roles influence employees' psychological states and behaviors. When AI functions as an “opponent”, it triggers job-related anxiety among employees. The fear of replacement can lead to negative behaviors such as service disruptions and knowledge hoarding. Conversely, if employees perceive this threat as a challenge, it can stimulate creativity. As an “assistant”, AI enhances work performance by reducing employees' cognitive load. Nevertheless, over-reliance on it may weaken employees' moral judgment and limit their access to diverse information. In the role of a “colleague”, AI collaborates with employees on complex tasks, but it also gives rise to issues like ambiguous job responsibilities and decreased interpersonal communication. When AI assumes the “leader” role, although its precise management has certain advantages, it also raises ethical concerns. Its impersonal communication style and potential privacy invasions can cause stress among employees, and algorithm-based management may undermine employees' creativity and job satisfaction. To address these impacts, the study proposes practical strategies for both organizations and individuals. At the organizational level, companies should provide AI-related training to help employees acquire the necessary skills and adapt to changes. Fostering a positive AI-friendly work culture can guide employees to utilize AI effectively and mitigate its negative effects. Additionally, establishing clear AI usage rules and ethical standards is essential for safeguarding employees' rights. At the individual level, employees need to actively develop skills that are difficult for AI to replicate, such as creativity and social skills, and continuously enhance their AI literacy. They should also shift their mindset towards AI, viewing it as an opportunity rather than a threat, to better integrate into the new workplace environment. Combining the current research status and the role-by-role analysis presented earlier, we also point out several directions for future research. First, we can further explore and refine the classification of AI in the Chinese context. Based on China's unique cultural, economic, and social environment, a more precise classification will provide a better understanding of how AI functions in Chinese workplaces. Next, we can explore the boundary conditions of such impacts. These boundary conditions include factors like organizational culture, employee characteristics, and technological maturity. Third, conducting comparative studies can reveal commonalities and differences in how AI affects employees in different contexts. This can be valuable for organizations operating globally or seeking to learn from different practices. Fourth, analyzing the differential impacts of AI across various industries and job functions helps understand its unique influences in different work settings. Different industries and job roles interact with AI in distinct ways, and this research can offer targeted insights. Fifth, conducting comparative studies can reveal commonalities and differences in how AI affects employees in different contexts. This can be valuable for organizations operating globally or seeking to learn from different practices. Finally, strengthening interdisciplinary research, which integrates knowledge from multiple fields like psychology, computer science, and management, offers a more comprehensive understanding. This study contributes to the existing literature by offering a comprehensive classification of AI roles in the workplace and elucidating their differential impacts on employees. By meticulously examining the four roles of AI in the workplace from the employees' perspective, it fills gaps in existing literature and offers a more comprehensive view of their interactions. It also provides practical insights for organizations and individuals to better adapt to the evolving role of AI in the workplace. Future research should further explore the boundary conditions of these impacts, investigate cross-cultural differences in employee responses to AI, and examine the long-term effects of AI on career development and work-life balance. Table and Figures | Reference | Related Articles | Metrics

Select

Investigating social cognitive characteristics of social anxiety within the Bayesian framework PENG Yujia, WANG Yuxi, JU Qianqian, LIU Feng, XU Jia Advances in Psychological Science    2025, 33 (8): 1267-1274.   DOI: 10.3724/SP.J.1042.2025.1267 Abstract （1841）   HTML （137）    PDF （1160KB）（1873）       Knowledge map    Social anxiety disorder (SAD) is among the most common anxiety disorders, marked by overwhelming fear and avoidance of social behaviors and social scenarios, and debilitates patients’ lives and work. Previous studies have provided ample evidence of dysregulated social cognition in social anxiety, such as negative cognitive biases, demonstrating a negative processing of social information. However, the factors driving the dysregulated social cognition remain unclear, impeding the elucidation of the underlying computational neural mechanisms of social anxiety symptoms and guiding personalized interventions. Within the Bayesian framework, the current project proposed that the negative cognitive biases phenomenon may stem from negative prior expectations. By integrating psychophysics experiments, electroencephalography (EEG), functional magnetic resonance imaging (fMRI), computational modeling, and machine learning, we will systematically investigate prior expectations' characteristics, formation, and dynamic modulation. The key innovation of this project lies in three major contributions. First, this study will be the first to propose and quantitatively examine the impact of prior expectations on dysregulated social cognition. Previous socially anxious research has primarily focused on behavioral manifestations and their associations with social information processing, yet largely overlooked the role of prior expectations in shaping social cognitive distortions. By identifying the static features and dynamic formation process of prior expectations in dysregulated social cognition, our study expands existing cognitive-behavioral models of social anxiety, providing a more comprehensive framework for understanding its underlying mechanisms. Second, our project aims to construct a mechanistic framework of social anxiety that systematically links behavioral manifestations to cognitive mechanisms, and further to neural mechanisms. By integrating behavioral experiments, neuroimaging, and computational modeling as methodological tools, we are able to map distorted cognitive components onto their specific neural underpinnings. This integrative approach provides robust empirical evidence, thereby advancing the theoretical understanding of social anxiety and offering a foundation for future research and intervention development. Third, this project extends the investigation to the translational level by evaluating the potential of neural decoding feedback as an intervention for social anxiety. By leveraging real-time neural data to modulate maladaptive social-cognitive expectations, we aim to assess the feasibility of neurofeedback-based treatments in social anxiety, providing a potential pathway for developing novel, data-driven therapeutic strategies. In summary, this project not only advances the theoretical understanding of social anxiety but also explores its translational potential. By extending the traditional cognitive-behavioral model to incorporate prior expectations and constructing a comprehensive behavioral-cognitive-neural framework, it systematically maps the progressive linkage from behavioral manifestations to cognitive processes and neural underpinnings—offering a novel perspective for studying anxiety-related disorders. Importantly, this project goes beyond theoretical contributions by identifying specific intervention targets derived from our computational framework and assessing their clinical applicability through neurofeedback. By leveraging real-time neural decoding to modulate maladaptive prior expectations, we aim to evaluate the efficacy of a novel, data-driven intervention approach. This translational effort holds promise for the development of precision-targeted treatments that can significantly enhance therapeutic outcomes for individuals with SAD. By elucidating the mechanisms underlying dysregulated social cognition through an integrative, multi-level approach, this project lays the foundation for a paradigm shift in both research and clinical practice. We encourage the broader adoption of computational psychiatry methods, redefine the understanding of dysregulated social cognition in social anxiety, and bridge the gap between mechanistic theory and personalized intervention. Ultimately, this work paves the way toward a new era of individualized, mechanism-informed mental health care empowered by technological innovation and theoretical precision. Table and Figures | Reference | Related Articles | Metrics

Select

A three-level meta-analysis of gender differences in spatial navigation ability XUE Xiaoran, CUI Wei, ZHANG Li Advances in Psychological Science    2025, 33 (5): 843-862.   DOI: 10.3724/SP.J.1042.2025.0843 Abstract （1801）   HTML （40）    PDF （1536KB）（2963）       Knowledge map    Spatial navigation refers to an individual's ability to update his or her position and orientation in space, learn the layout of new locations, and plan and follow routes to reach a destination. This ability is one of the fundamental abilities essential for the survival of both humans and animals. Gender is a significant factor contributing to individual differences in spatial navigation ability. Although numerous studies have explored gender differences in spatial navigation, findings regarding the existence and extent of these differences remain inconsistent. In these studies, gender often interacts with various factors, such as study design, collectively influencing spatial navigation ability. Therefore, it is essential to systematically investigate whether significant differences exist between men and women in spatial navigation ability and to analyze how moderating factors shape the relationship between gender and spatial navigation performance. The present study integrated 173 original papers involving 372 independent effect sizes and 26,604 subjects between 2007 and 2023 through a three-level meta-analysis to clarify gender differences in spatial navigation ability and their potential moderating variables. The results indicate that males outperform females in spatial navigation ability under most conditions and that this gender difference is significantly moderated by age, mode of representation, time constraints, task environment, test scenario, and assistive equipment. Specifically, age was a significant moderating variable: males significantly outperformed females in spatial navigation between the ages of 4 and 65 years, whereas there were no significant differences in infancy (0~4 years) and late adulthood (65 years and older). Representational modality in task design also played a role, with smaller gender differences in egocentric representational tasks and larger differences in allocentric representational tasks. Gender differences were more significant in time-constrained tasks and less so when there were no time constraints. Task environment and test conditions also significantly affected results, with smaller gender differences in indoor environments only, real-scene tests, or conditions using paper-and-pencil instruments and no assistive devices, and more significant differences in dual indoor-outdoor test conditions or conditions using assistive devices. In addition, the study found that geographic region did not influence gender differences in spatial navigation ability, as participants from various continents consistently exhibited significant gender differences. On the one hand, this may be due to the uneven focus on different regions in previous research, resulting in an imbalanced distribution of participants. On the other hand, economic conditions and living environments may serve as more proximal factors influencing gender differences, potentially moderating these differences by providing varying resources and challenges. Furthermore, gender differences in spatial navigation ability remained consistent across different task types and measurement metrics, indicating the reliability of these assessments. However, in real-life situations, spatial environmental cues are considerably more complex and multifaceted than what simplified tasks in laboratory settings can fully capture. Therefore, future research should continue to explore evaluation methods that are more closely aligned with real-world conditions. This study examined gender differences in spatial navigation ability and its moderators through a three-level meta-analysis. The findings not only confirmed the phenomenon of male superiority in spatial navigation ability but also identified several key variables that moderate these gender differences. Future research should involve more diverse and representative population samples and employ task designs and measurement methods that closely reflect real-world environments to further explore the relationship between gender and spatial navigation ability. Moreover, in practical terms, educators should prioritize the development of students' spatial navigation skills, particularly through targeted teaching and hands-on activities aimed at enhancing women's confidence and ability in handling navigation tasks. These efforts will not only contribute to narrowing the gender gap but also improve students' ability to tackle spatial challenges in real-world contexts, ultimately promoting educational equity and social progress. Table and Figures | Reference | Related Articles | Metrics

Select

Empathy in large language models: Evaluation, enhancement, and challenges ZHOU Qianyi, CAI Yaqi, ZHANG Ya Advances in Psychological Science    2025, 33 (10): 1783-1793.   DOI: 10.3724/SP.J.1042.2025.1783 Abstract （1700）   HTML （126）    PDF （535KB）（4378）       Knowledge map    Amid the rapid evolution of artificial intelligence technologies, the application scope of large language models (LLMs) has extended beyond traditional information processing tasks to novel domains involving the simulation of complex human emotions and interactions. Particularly in emotion-intensive contexts such as psychological counseling, physician-patient communication, and customer service, the capacity of LLMs for empathy simulation has emerged as a focal point in academic research and demonstrates substantial potential for real-world application. However, fundamental questions remain: What are the essential differences between LLM-simulated empathy and human empathy? How can we evaluate such capabilities in a scientific and comprehensive manner? What is the current state of development, and what are the core bottlenecks? More critically, how can LLMs’ empathetic performance be effectively enhanced while addressing the associated ethical risks? While existing studies have explored some of these issues, a systematic integrative framework is still lacking. Therefore, this study conducts a comprehensive analysis of empathy simulation in LLMs across four key dimensions: evaluation methods, current development status, enhancement strategies, and critical challenges. The goal is to provide a theoretical foundation and directional guidance for future research and practical deployment in this domain. Currently, the evaluation of LLMs’ empathy simulation can be categorized into three main approaches: human-based, automated, and task-driven. Human evaluation relies on subjective ratings or comparative judgments made by human annotators or domain experts, and excels at capturing nuanced emotional perceptions and context-dependent subtleties. However, it suffers from high subjectivity and cost. Automated evaluation employs computational techniques such as sentiment classification and cosine similarity for objective quantification, offering efficiency and reproducibility suitable for large-scale testing. Nonetheless, it often fails to account for contextual or subtle emotional variations and cannot adequately assess the naturalness or perceived empathy of language. Task-driven evaluation involves designing specific tasks such as emotion cause recognition (e.g., RECCON) or leveraging psychological empathy paradigms and standardized scales (e.g., IRI, BES) to assess model performance. This method aligns more closely with real-world applications and yields quantifiable metrics, though its generalizability is constrained by the specific design of tasks. This study compares the strengths and limitations of the three approaches and highlights the lack of a unified evaluation framework. It emphasizes the urgent need to develop a standardized and integrated assessment system, particularly one that incorporates psychological measurement paradigms to probe deeper empathetic response mechanisms within LLMs. Recent studies using these varied evaluation methods suggest that LLMs can generate empathetic responses comparable to, or even surpassing, human outputs in certain scenarios, providing effective emotional support. However, there is still considerable room for improvement in specific empathy-related subtasks and in performance on standardized empathy scales—especially when confronting complex or mixed emotions. To enhance empathy simulation in LLMs, four key strategies are proposed: data augmentation, architectural and framework optimization, reinforcement learning, and prompt engineering. Data augmentation involves constructing larger, higher-quality, and culturally diverse empathetic dialogue datasets for fine-tuning. Architectural and framework optimization refers to the development of novel model structures capable of dynamically capturing emotional and personality traits (e.g., Pecer), or hybrid frameworks combining expert models and chain-of-thought reasoning (e.g., HEF, EBG) to improve the understanding of fine-grained emotions. Reinforcement learning integrates feedback from humans or other AI agents (e.g., the Muffin framework) to reward high-quality responses and guide the generation of empathy-aligned outputs. Prompt engineering can embed psychological theories—such as cognitive behavioral therapy—into prompt design, guiding the model to conduct deeper emotional reasoning and generate contextually appropriate responses from the input level. Nevertheless, LLMs face persistent and, in some cases, insurmountable technical limitations. Their empathy simulation is fundamentally rooted in large-scale statistical pattern matching, rather than genuine emotional experience or intrinsic motivation. As a result, their responses often appear formulaic and lack authenticity. Moreover, LLMs struggle with interpreting complex emotions, sarcasm, and culturally variable expressions of empathy. Their adaptability to diverse cultural contexts and emotionally ambiguous situations remains limited. In addition, the use of LLMs inevitably raises ethical concerns, including the potential generation of harmful or discriminatory content, misuse for information manipulation, and the risk of users developing excessive emotional dependence on AI systems—potentially undermining real-life social interactions. In conclusion, this study presents a comprehensive investigation into LLMs’ empathy simulation from the perspectives of evaluation, current capabilities, enhancement strategies, and inherent challenges. It identifies key issues and delineates future research directions, offering a conceptual foundation for advancing both academic inquiry and practical implementation in this emerging field. Table and Figures | Reference | Related Articles | Metrics

Select

Marketing effect of virtual influencers and its mechanisms in the context of AI technology LI Yan, CHEN Wei, WU Ruijuan Advances in Psychological Science    2025, 33 (8): 1425-1442.   DOI: 10.3724/SP.J.1042.2025.1425 Abstract （1654）   HTML （81）    PDF （769KB）（1149）       Knowledge map    With the rapid advancement of AI technology, virtual influencers have become an increasingly prominent presence on social media. These AI-driven digital personas not only attract attention by sharing engaging content but also directly promote and offer products or services to their followers, facilitating transactions and exerting a profound influence on consumer behavior. As a result, an increasing number of brands are actively leveraging virtual influencers to enhance brand visibility, promote products, and establish deeper connections with their audiences. The application of virtual influencers in the advertising and marketing industry has already shown promising results, highlighting their potential as an effective tool for consumer engagement. Powered by AI technology, virtual influencers can realistically mimic human characteristics and personalities, enabling them to precisely appeal to specific market segments. By meticulously designing their appearance, behavior, and communication style, these virtual entities create a strong sense of relatability, effectively engaging audiences. As an emerging marketing approach, virtual influencer marketing has already demonstrated considerable effectiveness in practice. However, despite its increasing adoption, theoretical research on this phenomenon remains in its early stages, necessitating a systematic review of existing studies to assess its current research status and development trends. To this end, this study first clarifies and defines the concept and connotation of virtual influencers. At the character design level, it identifies several key factors that influence the marketing effectiveness of virtual influencers, including character backstory, distinctive personality, emotional module, and controlling entity. These elements play a crucial role in shaping consumer perceptions and engagement levels. Furthermore, based on the two dimensions of form realism and behavioral realism, this study innovatively categorizes virtual influencers into six distinct types: spokesperson humanlike virtual influencers, influencer humanlike virtual influencers, spokesperson anime-like virtual influencers, idol anime-like virtual influencers, mascot nonhumanlike virtual influencers, and storyteller nonhumanlike virtual influencers. The study further investigates the mechanisms and moderating factors that contribute to both the positive and negative marketing effects of virtual influencers. Compared to human influencers, virtual influencers present significant advantages that can lead to more favorable marketing outcomes for businesses: (1) Companies can maintain complete control over the behavior and performance of virtual influencers, ensuring a high degree of brand alignment; (2) Virtual influencers can effectively capture consumers’ attention and create a sense of novelty, thereby enhancing consumers’ perception of advertising and brand innovation; (3) The use of virtual influencers allows brands to mitigate risks associated with human endorsers, such as controversies, scandals, or reputational damage, thereby ensuring greater stability in long-term marketing campaigns. However, the use of virtual influencers may also lead to certain negative effects: (1) Algorithm aversion may cause consumers to resist AI-generated virtual digital humans, particularly when they have encountered algorithmic errors, unnatural interactions, or unsatisfactory results; (2) According to the uncanny-valley effect, virtual influencers that appear excessively human-like may evoke discomfort or negative psychological reactions among consumers; (3) Consumers may develop an awareness of falsity of these virtual influencers, leading to distrust, which in turn negatively impacts their credibility and endorsement effectiveness. Furthermore, this study examines several crucial moderating factors from both the advertising design and consumer behavior perspectives. These factors include source transparency (i.e., whether the AI-generated nature of the virtual influencer is disclosed), product categories and characteristics, application scenarios, and individual differences among consumers. Finally, this study summarizes the marketing effects of virtual influencers and explores future research directions in key areas such as technological empowerment, underlying mechanisms, marketing outcomes, and ethical considerations. Additionally, it examines the broad development prospects of virtual influencer marketing, including applications in advertising endorsements, brand personification, and customer engagement. This study also identifies critical challenges, such as technological and cost barriers, market acceptance, content management, and legal and ethical concerns. This study not only contributes to the academic understanding of virtual influencer marketing but also provides practical guidance and recommendations for businesses and marketing practitioners regarding the application and strategic planning of virtual influencers. Table and Figures | Reference | Related Articles | Metrics

Select

Trends and dynamics in longitudinal research: Model development, integration, and differentiation LIU Yuan, YAO Zhichen Advances in Psychological Science    2025, 33 (7): 1181-1198.   DOI: 10.3724/SP.J.1042.2025.1181 Abstract （1594）   HTML （38）    PDF （1033KB）（5709）       Knowledge map    “Trends” and “dynamics” represent two significant themes in longitudinal research. To address the challenges associated with trends and dynamics, researchers frequently utilize panel data or intensive longitudinal data to gather information, allowing for the development of various statistical models. The trends study analyzes systematic changes and usually operates at the between-person level. It illustrates the general developmental trajectory while acknowledging individual differences. To effectively capture these general trends, researchers often use panel data, as it involves wide intervals and prolonged periods of data collection. A commonly employed model for this purpose is the latent growth model (LGM) combined with a multilevel model (MLM). In contrast, the dynamic study focuses on temporal changes within individuals, typically at the within-person level. It examines autoregressive and cross-lagged relationships from earlier time points to later ones. Both panel data and intensive longitudinal data could address dynamic issues because they allow for the measurement wave over short time intervals and necessitate a large number of measurements. Cross-lagged panel models (CLPM) are frequently used to analyze dynamics for panel data, while time series analysis and dynamic structural equation models (DSEM) are commonly applied to intensive longitudinal data. As research questions become more intricate, we rely on models that integrate trends and dynamics, yielding numerous integrated models. For panel data, this includes the random intercept cross-lagged model (RI-CLPM), autoregressive latent trajectory model (ALT), and latent curve model with structural residuals (LCM-SR), among others, to provide a comprehensive approach to addressing the interplay of trends and dynamics. In the context of intensive longitudinal data, where stationary is a preassumption for time series modeling, models incorporating the detrending process have been developed, such as the dynamic structural equation model (DSEM) and residual dynamic equation model (RDSEM), etc. We utilized empirical data from the 2013 Health and Retirement Study (HRS) to demonstrate the practical application of various longitudinal models. Our findings revealed that the integrated models, such as ALT and LCM-SR, exhibited superior model fit. This suggests that there are developmental trends that must be accounted for. The ALT model displayed significant autoregressive and cross-lagged relationships among the target variables, whereas the LCM-SR models did not. In conclusion, we compared various longitudinal models and provided practical recommendations. First, researchers should determine the appropriate data collection paradigm to employ. When the number of measurement waves is ten or fewer and the time intervals are large, panel data is suitable; otherwise, an alternative approach should be considered. The long format is also suggested for intensive longitudinal data. Second, since the stationary is crucial in dynamic research, it is essential to assess trends. Panel data can be analyzed for trends using LGM or MLM with time covariates, while intensive longitudinal data (through time series analysis) should employ stationary tests. Descriptive statistics can also provide valuable insights. If trends are present, panel data should utilize an integrated model that encompasses both trends and dynamics, whereas intensive longitudinal data should adopt detrending models. In the absence of trends, direct dynamic modeling can be applied. Specifically, if the goal is to distinguish between trends and dynamics, researchers should consider residual models such as RI-CLPM, LCM-SR, and RDSEM. Conversely, for research emphasizing dynamics, cumulative models like ALT and DSEM should be applied. Table and Figures | Reference | Related Articles | Metrics

Select

Unconscious fear and its neural mechanisms YU Lingfeng, ZHANG Jie, MING Xianchao, LEI Yi Advances in Psychological Science    2025, 33 (7): 1234-1245.   DOI: 10.3724/SP.J.1042.2025.1234 Abstract （1583）   HTML （60）    PDF （726KB）（2178）       Knowledge map    Unconscious fear refers to fear responses that occur without conscious awareness of fear-inducing stimuli. Traditional views suggest that unconscious fear processing primarily relies on the superior colliculus-pulvinar-amygdala pathway, with the amygdala serving as the central hub. However, recent studies on cortical and subcortical structures have significantly expanded this understanding. Research has revealed that beyond the amygdala, structures including the visual cortex, lateral geniculate nucleus (LGN), superior colliculus, and pulvinar can rapidly detect and specifically respond to unconscious fear stimuli, revealing a more complex neural processing network. While numerous studies have explored the neural basis of unconscious fear processing, a systematic integration and analysis of the roles and interactions of various brain regions during early visual processing stages remains lacking. Therefore, this paper focuses on examining the functions and interactions of visual cortical and subcortical regions (LGN, superior colliculus, pulvinar) during unconscious fear processing, aiming to construct a more comprehensive neural network model. The LGN's function extends far beyond its traditional role as a simple visual relay station. Research indicates that K cells in the LGN project to the visual cortex, forming the neural basis for blindsight. More importantly, the extensive connections between the LGN, thalamic reticular nucleus, and visual cortex enable selective filtering of visual information before conscious processing. Under threat conditions, signal transmission rates from the retina to LGN significantly increase, while the ventral LGN participates in modulating the duration of defensive responses to visual threats. Furthermore, studies on blindsight patients further confirm that the LGN can process threat signals bypassing V1, with functional connections between the LGN and visual cortex providing an alternative pathway for rapid behavioral responses, highlighting the LGN's importance as a key node in unconscious fear processing. The pulvinar's role in unconscious fear processing is more sophisticated than traditionally understood. While conventional views suggest that the pulvinar merely relays information directly to the amygdala, research has revealed more refined functional divisions, where the inferior pulvinar primarily connects with extrastriate visual areas and superior colliculus, while the medial pulvinar maintains bidirectional connections with the amygdala and frontoparietal regions. Notably, the medial pulvinar's approximately 200ms response latency suggests that it may participate in coordinating cortical assessment of stimulus significance before information reaches the amygdala, rather than simply relaying signals. This complex connectivity pattern establishes the pulvinar as a crucial coordinator in evaluating stimulus biological significance. The primary visual cortex demonstrates unique capabilities in unconscious threat processing. Through feedforward sweep mechanisms, V1 can rapidly process fear signals independently of feedback from higher visual areas. More significantly, V1 undergoes plasticity through fear learning, forming threat-related memory representations that facilitate rapid and precise processing under unconscious conditions. This plasticity is influenced by multiple regulatory mechanisms, including enhanced amygdala theta oscillations, increased acetylcholine release from the basal forebrain, and enhanced signal transmission between the amygdala and sensory cortex. Future research should focus on three key directions: First, employing multimodal MEG/EEG-fMRI imaging techniques combined with dynamic causal modeling and Granger causality analysis to investigate temporal characteristics and dynamic interactions between cortical and subcortical structures; Second, developing large-scale neural network computational models to simulate dynamic interactions between key brain regions and predict new neural circuit interaction patterns; Finally, exploring clinical translational applications, particularly developing novel treatment approaches based on unconscious exposure therapy and neurofeedback training, providing new therapeutic strategies for emotional regulation disorders and mental illnesses. In conclusion, this paper establishes that alongside the amygdala, the LGN, pulvinar, and primary visual cortex constitute critical neural nodes in unconscious fear processing. This not only reflects the diversity of visual processing mechanisms but also exemplifies the distinctive neural patterns that characterize human threat response systems. Table and Figures | Reference | Related Articles | Metrics

Select

Exploring the impact of focus back effort training on mind wandering and its mechanisms HE Hong, ZHANG Xinyue, SHI Jinghong, LIU Qiang Advances in Psychological Science    2025, 33 (7): 1077-1090.   DOI: 10.3724/SP.J.1042.2025.1077 Abstract （1541）   HTML （80）    PDF （1791KB）（1775）       Knowledge map    Humans spend a significant amount of time engaged in mind wandering. While it is a common cognitive state, it is often seen as a hindrance to productivity and daily functioning. Minimizing its frequency is crucial for maintaining focus and efficiency in academic, professional, and personal life. This study is grounded in the resource control theory of mind wandering and the dynamic framework model, proposing that focus back effort reflects executive control and the intentional constraint of thought. Specifically, resource control theory emphasizes the role of executive control in reducing mind wandering. Focus back effort training enhances this process by lowering the depletion of executive control resources, thereby minimizing the extent to which mind wandering consumes task-related resources. The dynamic framework model, on the other hand, highlights both automatic and intentional constraints on thought. Focus back effort represents intentional constraint, and training reduces the reliance on such deliberate control, making thoughts more likely to automatically return to the task at hand, ultimately decreasing mind wandering. The method of using focus back effort training to reduce the frequency of mind wandering is beneficial to make up for the resource consumption caused by the de-automation of mindfulness training. From the perspective of cognitive intervention, combined with brain imaging methods, this paper intends to conduct research in the following three aspects. Firstly, behavioral experiments and task-state magnetic resonance imaging data are used in the laboratory to explore the effectiveness of the intervention of focus back effort training on mind wandering and task performance and to reveal its cognitive neural mechanism. Secondly, the laboratory research was extended to life situations to investigate the effects of focus back effort training in life situation. Finally, the research explores the effect of focus back effort training on classroom mind wandering and academic performance of primary and middle school students. Resource control theory focuses on the irrelevance of task-unrelated content, while the dynamic framework model emphasizes the fluid transition between different mental contents. This study proposes that focus back effort embodies key elements of both theories: executive control from resource control theory and thought constraint from the dynamic framework model. Both executive control and thought constraint are considered higher-order structures of mind wandering. Investigating the effects of focus back effort training on mind wandering can help validate the hierarchical nature of these two theories. Furthermore, by employing multimodal research methods to effectively reduce the frequency of mind wandering across different contexts, this study not only provides robust empirical support for and extends these theoretical frameworks but also facilitates the integration of static and dynamic perspectives. Ultimately, this contributes to a more comprehensive understanding of the mechanisms underlying mind wandering. Beyond these contributions, this study presents several key innovations: First, it pioneers the intervention of mind wandering through psychological factors, a novel approach proposed by our research team. While previous studies have primarily focused on correlational findings, this study advances the field by developing a focus back effort training paradigm and conducting longitudinal intervention research to establish causal relationships, thereby extending prior work. Second, it integrates behavioral experiments with functional magnetic resonance imaging studies, conducting research both in controlled laboratory settings and ecologically valid real-life contexts. This dual approach ensures that the findings are not only scientifically rigorous but also broadly applicable, offering new insights for future research in mind wandering and consciousness. Finally, as mind wandering is a pervasive phenomenon in daily life, exploring its cognitive and neural mechanisms in real-world contexts provides valuable strategies for mitigating its negative impact on work and daily functioning. Moreover, applying focus back effort training in classrooms to reduce mind wandering among primary and secondary school students and enhance academic performance has significant educational implications. Table and Figures | Reference | Related Articles | Metrics

Select

Dialect stereotypes in advertising: Effects and theoretical mechanism LING Bin, LIU Yingkai Advances in Psychological Science    2025, 33 (8): 1408-1424.   DOI: 10.3724/SP.J.1042.2025.1408 Abstract （1515）   HTML （65）    PDF （676KB）（997）       Knowledge map    In the current social context characterized by the coexistence of globalization and multiculturalism, dialects, as linguistic variants, play a significant role in advertising and marketing. Dialect stereotypes refer to people's fixed and generalized cognitive evaluations of dialects, which are inherently complex and multidimensional, unconsciously influencing individuals' attitudes and behaviors toward dialect advertisements. There persists a systematic gap in addressing three critical questions: (1) What is the concrete meaning of dialect stereotypes, (2) how dialect stereotypes affect the persuasive effectiveness of advertisements across different theoretical perspectives, and (3) what specific boundary conditions moderate these effects. First, this paper commences by precisely defining dialects and related concepts such as accents, colloquialisms, jargon, internet slang, and phonetic variations. Next, it systematically elaborates dialect stereotypes through three core dimensions: (a) linguistic features that cover phonological patterns, lexical choices, phrasal structures, and pragmatic conventions; (b) user image constituted by status, solidarity, and dynamism; and (c) social culture manifested through regional identity shaping, emotional resonance intensification, local culture representation, and character persona formation, which collectively delineate the research scope of dialects and their stereotypes in advertising. Then, building upon four theoretical frameworks—markedness theory, dual process theory, social identity theory, and spatial presence—the paper thoroughly explores the underlying mechanisms of dialect stereotype effects in advertising and reveals hierarchical differences in explanatory level and focus. Specifically, markedness theory, rooted in dialect linguistic features, emphasizes how the markedness of dialect advertisements influences consumer expectation levels, with particular attention to advertisement comprehensibility and memorability. Dual process theory focuses on speaker image, detailing the distinct roles of system 1 (automatic processing) and system 2 (controlled processing) in enhancing brand awareness and advertisement credibility. Social identity theory addresses the cultural connotations of dialects and users’ identity, highlighting cognitive and affective responses such as in-group favoritism and out-group discrimination. Spatial presence concerns how dialect linguistic and cultural attributes synergistically create immersive experiences that shape product authenticity perceptions and consumer experience. Concurrently, the paper clarifies the explanatory boundaries of these theories and analyzes five moderating factors: consumers' individual traits, product attributes, brand characteristics, spokespersons' individual traits, and advertising appeals. By delineating theoretical distinctions and interconnections, this paper identifies distinct cognitive processing pathways associated with each theory. Markedness theory plays a pivotal role during initial information processing, demonstrating how dialect advertisements rapidly capture consumer attention through distinctive linguistic features. Dual process theory further refines subsequent processing stages: System 1 triggers automated processing based on preexisting dialect stereotypes, eliciting social identity effects, while System 2 engages consumers in deliberate evaluation of advertisement content. At this stage, Spatial presence explains how dialect linguistic characteristics and cultural attributes enhance product authenticity perceptions and enrich consumers’ sensory experiences. Given the potential simultaneity of these interrelated cognitive processes, the paper proposes integrative possibilities across theoretical frameworks: System 1 processing in dual-process theory corresponds to stereotype-based cognitive-affective responses encompassing social identity effects, whereas markedness theory implicitly incorporates fluency processing of dialect information when explaining advertisement markedness. These theoretical intersections provide novel directions for future research. At the same time, the paper further advances a theoretical model of dialect stereotype effects in advertising, which can offer some implications for advancing theoretical frameworks and mechanistic investigations in dialect advertising research. Future research should prioritize three directions. First, diversifying linguistic forms in dialect advertisements through innovative approaches such as dialect-standard language hybrid advertisements and specific dialectal features (e.g., inverted sentence structures, retroflex suffixes, reduplication patterns). Second, examining how cultural values—particularly collectivism-individualism orientations, temporal perspectives, and cultural identification levels—moderate advertising persuasion effects through empirical validation of underlying mechanisms. Third, exploring artificial intelligence (AI) applications in dialect advertising, including investigating the effectiveness of dialect advertisements in AI environments, identifying key influencing factors, and developing theoretical foundations for integrating dialect advertising design with AI technologies to drive innovation in advertising practices. Table and Figures | Reference | Related Articles | Metrics

Select

The impact of brand-developing versus collaborative virtual influencer endorsement selection strategies on consumer engagement XIANG Diandian, YIN Yule, GE Mengqi, WANG Zihan Advances in Psychological Science    2025, 33 (6): 965-983.   DOI: 10.3724/SP.J.1042.2025.0965 Abstract （1503）   HTML （48）    PDF （715KB）（3967）       Knowledge map    Virtual influencer endorsements, empowered by artificial intelligence technology, offer a new approach to brand communication and exert outstanding influence on social media platforms. However, research on the relationship between virtual influencer endorsements and consumer engagement remains in its early stages, not only limited in scope but also facing two primary research gaps. First, firms face a unique challenge in leveraging these endorsements to enhance consumer engagement. This challenge arises from a strategic tension in selecting either brand-developing proprietary virtual influencers or collaborating with third-party virtual influencers. However, there is a lack of exploration of the impact of virtual influencer endorsement selection strategies on consumer engagement. Second, existing literature has yet to investigate the underlying mechanisms and boundary conditions. From various theoretical perspectives, virtual influencer endorsement involves several key mechanisms, including the communication cycle, information processing, and relationship interaction. Therefore, understanding the effectiveness of these mechanisms is crucial for addressing the differing brand communication effects generated by different virtual influencer endorsement selection strategies. In response to these research gaps, based on the communication time, accuracy, and relationships perspectives, this study proposes communication timeliness mechanisms (Study 1), information discrimination mechanisms (Study 2), and relationship-building mechanisms (Study 3) to comprehensively examine the effects and mechanisms of virtual influencer endorsement selection strategies (brand-developing proprietary versus collaborating with third-party) on consumer engagement, and explores key boundary conditions from artificial intelligence technology factors and virtual influencer/consumer factors perspectives, combined with brand communication-related theories and influencer marketing. Specifically, first, drawing on the research insights on brand awareness in the process of brand communication in marketing, and based on the signal detection theory, this study aims to explore the communication time mechanism of long-term "brand awareness" and "virtual influencer awareness" induced by virtual influencer endorsements selection strategies, and further analyze the moderating effect of the artificial intelligence technology factor (i.e., authenticity level of virtual influencers) and influencer factor (i.e., brand-virtual influencer consistency) (Study 1). Second, grounded in the elaboration likelihood model (ELM), which is widely used in the advertising domain, this study aims to explore the double-edged sword information discrimination mechanism of advertising intrusiveness and information diagnostic induced by virtual influencer endorsements selection strategies selection and examine the moderating effect of artificial intelligence technology factor (i.e., authenticity level of virtual influencers) and consumer factor (i.e., need for cognition) (Study 2). By doing so, it comprehensively understands how consumers distinguish the endorsement information of virtual influencers. Third, following the widely used parasocial interaction theory in the field of social media marketing, this study explores the relationship-building mechanism of consumers in virtual influencer endorsement brand activities, that is, consumer-brand identification and consumer-virtual influencer identification, and examines the moderating effect of artificial intelligence technology factor (i.e., virtual influencer's agency: fully artificial intelligence-driven versus manual intervention) the authenticity level of virtual influencers and influencer factor (i.e., brand-virtual influencer consistency) (Study 3). In response to these research gaps, this work makes the following contributions. First, this article focuses on the uniqueness of virtual influencer endorsement selection strategies, thus expanding the scope of brand endorsement research. While existing studies primarily explore the differences between virtual and real influencers in brand communication, few studies examine the specific selection strategies for virtual influencer endorsements. By analyzing this selection strategy, this study enriches the theoretical research system of brand endorsement and provides new topics for further research in this field. Additionally, this study systematically investigates the multiple mediation models, comprehensively revealing the underlying mechanisms, which contributes to understanding the relationship between virtual influencer endorsements selection strategies and consumer engagement and advancing virtual influencer marketing research. Furthermore, this study reveals the boundary conditions related to artificial intelligence technology factors (i.e., authenticity level of virtual influencers, virtual influencer's agency: fully artificial intelligence-driven versus manual intervention), virtual influencer factor (i.e., brand-virtual influencer consistency), and consumer factor (i.e., need for cognition), thereby facilitating interdisciplinary integration between the fields of influencer marketing, AI technology and psychology. Moreover, this study aims to offer effective guidance for firms in selecting virtual influencer spokespersons, managing interventions for virtual influencer applicability, and implementing artificial intelligence marketing, thereby providing crucial insights for virtual influencer practices. Table and Figures | Reference | Related Articles | Metrics

Select

Neurophysiological mechanisms and interventions of subthreshold depression by integrating machine learning techniques LIU Yongjin, YANG Xue, DU Xinxin, JI Wenqi, ZANG Yinyin, GUAN Ruiyuan, SONG Sen, QIAN Mingyi, MU Wenting Advances in Psychological Science    2025, 33 (6): 887-904.   DOI: 10.3724/SP.J.1042.2025.0887 Abstract （1492）   HTML （70）    PDF （1298KB）（4913）       Knowledge map    Major Depressive Disorder (MDD) is a significant global public health threat for both national mental health, causing significant societal burden for both individuals and families. Subthreshold depression (StD) is an ultra-high-risk phase for MDD. Understanding its neurophysiological mechanisms and dynamic development patterns can help predict MDD onset and inform the development of novel and effective preventive interventions. To overcome the limitation of regarding depression as a static and single diagnosis, this study uses dynamical systems theory and machine learning techniques to explore the StD's neurophysiological mechanisms. By combining resting-state, task-state evaluations, and clinical interviews, it constructs a neurodynamic network model for StD with multi-modal data and machine learning, followed by cross-population and cross-individual validation. Firstly, using multi-modal data, Study 1 explores the differences in symptoms and neurophysiological mechanisms among individuals with StD, healthy, and MDD individuals from a population perspective, and then, identifies critical predictive neurophysiological indicators of StD. By combining resting-state and task-state evaluations, as well as clinical interviews, it explores the distinctiveness of symptoms and multi-modal data in StD individuals (versus healthy and MDD). And the neurodynamic network model for StD will be constructed using multi-modal data combined with machine learning techniques to analyze depressive symptoms and neurophysiological characteristics, followed by cross-population and cross-individual validation. at the individual level, Study 2 use ecological momentary assessment combined with follow-up tracking to investigate the relationship between the attractor state of subthreshold depression and subsequent transition to MDD, along with accompanying neurophysiological changes. A dynamic development prediction model will be constructed. This study utilizes longitudinal tracking of multi-modal data combined with machine learning techniques to measure the attractor of StD and explore its predictive ability for subsequent depressive states and neurophysiological characteristic changes. Finally, to overcome the limitations of evaluating the preventive intervention effects for StD mainly relied on clinical diagnostic results, this study explores the effectiveness of cognitive behavioral therapy (CBT) in preventing the progression of StD from a commonality perspective, with the depressive symptoms and neurophysiological characteristics as objective evaluation indicators of intervention effectiveness. Furthermore, it elucidates the predictive role of the individual's attractor state in the future transformation of subthreshold depression. Study 3 uses a randomized controlled trial to investigate the preventive intervention efficacy of CBT group (versus waiting list group) on subthreshold depression and the mediating role of the attractor state. Based on the above study, this study proposes a predictive hypothesis regarding the preventive effect of cognitive behavioral therapy (CBT) on subthreshold depression based on the dynamical systems theory. It is hypothesized that CBT can reduce depressive symptoms and the transition rate by enhancing the stability of the attractor state, and this effect can be measured by neurophysiology characteristics. In conclusion, this study offers a dynamic-systems-theory-driven approach, integrating multi-modal machine learning techniques to analyze the neurophysiological uniqueness of StD. This study helps develop more precise models for predicting StD and evaluating preventive intervention effects based on neurophysiological characteristics obtained from various sensors. These results can be used for daily physiological assessments, early detection of emotional or mental states earlier deterioration, and development of more precise early diagnosis and prevention strategies for MDD. Table and Figures | Reference | Related Articles | Metrics

Select

Memory consolidation during wakeful rest: Evidence from EEG and fMRI LEI Xu, WENG Linman, YU Jing Advances in Psychological Science    2025, 33 (5): 729-743.   DOI: 10.3724/SP.J.1042.2025.0729 Abstract （1414）   HTML （68）    PDF （983KB）（4038）       Knowledge map    Both wakeful rest and sleep are beneficial for offline memory consolidation. However, our understanding of the connections and differences in memory consolidation between these two states, particularly regarding the shared cognitive neural mechanisms, remains limited. This study will focus on “memory consolidation during wakefulness”, using declarative and procedural memory tasks to examine memory consolidation activities under natural conditions as well as during modulation by neural replay-based closed-loop Targeted Memory Reactivation (TMR) and closed-loop electrical stimulation. The aim is to investigate the roles of wakeful rest and sleep in memory consolidation and explore the underlying neural mechanisms involved. To this end, the study will address the following key objectives: (1) Propose a unified theory of offline memory consolidation that spans both sleep and wake states, leveraging the identification of common characteristics between these states as a breakthrough to explore neural biomarkers of memory consolidation during wakeful rest. (2) Use neural replay activity as an entry point, this study will capture it to pinpoint the time window during which memory consolidation occurs and specifically identify the relevant neural features. (3) Building on sleep-state research to verify the effectiveness of the neural replay-based closed-loop TMR and provide guidance for its application during wakeful rest, while exploring the corresponding neural mechanisms. (4) Investigate the modulatory effects of direct hippocampal stimulation on memory consolidation and develop an electrical stimulation protocol for memory enhancement. (5) Conduct long-term follow-up studies to assess the effects of memory consolidation interventions over time, observing changes in memory performance across extended time scales, verifying the ecological validity of the interventions, and exploring the potential to apply laboratory findings to real-world learning contexts. This study presents three major innovations. First, it enhances our understanding of the role of wakeful rest in facilitating memory consolidation. While sleep has been extensively studied in the context of offline memory consolidation, with its mechanisms well understood, research on memory consolidation during wakefulness remains insufficient and requires further in-depth exploration. Currently, most human studies focus on the behavioral level, with few examining the underlying neural mechanisms, which limits our understanding of offline memory consolidation during wakefulness. Therefore, this study specifically focuses on memory consolidation during wakefulness and conducts a series of experiments to broaden our understanding of offline consolidation. Second, it offers an accurate characterization of the macroscopic neural representation of offline memory consolidation during wakeful rest. Neural replay, a key mechanism in memory consolidation, is challenging to detect directly in healthy humans using non-invasive methods. However, with the aid of computational neuroscience techniques, we can capture neural replay activity using non-invasive neuroimaging techniques, such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI). By focusing on neural replay activity, this study offers a more precise depiction of the neural processes involved in offline memory consolidation during wakeful rest, in contrast to traditional approaches that rely on correlation analysis to infer neural representations. Third, it provides new approaches to memory regulation. This study leverages neural replay activity to explore closed-loop TMR and closed-loop electrical stimulation as novel memory regulation techniques. By providing new insights into memory modulation, this study offers valuable directions for future research on memory intervention. In summary, this study aims to utilize advanced techniques such as EEG, fMRI, temporal interference (TI) electrical stimulation, and computational neuroscience techniques to capture the dynamic memory consolidation activities during both waking and sleep states, uncover the core characteristics of offline memory consolidation, and explore novel pathway for memory regulation based on real-time neural feedback. The implementation of this study will be instrumental in elucidating the neural mechanisms underlying memory consolidation across different brain states and laying the foundation for regulating memory consolidation during wakeful rest. Going forward, this study aims to apply its findings to educational interventions, such as learning strategy design, and initiate translational research to unlock the full potential of these interventions in real-world applications. Table and Figures | Reference | Related Articles | Metrics

Select

The inhibitory effects of exercise intervention on aggressive behavior and its mechanisms CHU Kequn, ZHU Fengshu Advances in Psychological Science    2025, 33 (7): 1257-1266.   DOI: 10.3724/SP.J.1042.2025.1257 Abstract （1388）   HTML （63）    PDF （461KB）（4187）       Knowledge map    This paper explores the multidimensional pathways and influencing factors of exercise intervention in suppressing aggressive behavior. By integrating four theoretical frameworks—emotional regulation, cognitive control, social connection, and social identity recognition—we propose a dynamic interactive comprehensive model. This model systematically elucidates how exercise interventions mitigate aggressive behaviors through improving emotional states, enhancing cognitive functions, reinforcing behavioral synchrony, and boosting group identity. The study further analyzes the modulation effects of intervention types, cultural backgrounds, and individual characteristics. Additionally, we discuss the neurobiological and physiological mechanisms supporting these theoretical pathways and suggest that future research may optimize model validation through dynamic modeling techniques and other innovative methods. Numerous empirical studies have supported the significant effectiveness of exercise interventions in reducing different types of aggressive behaviors, both in the short and long term. For instance, a 12-week aerobic exercise program demonstrated a marked reduction in verbal and physical aggression among adolescents. Acute high-intensity exercise has also shown positive effects in reducing aggressive scores following conflict simulation tasks. While both individual and group exercises contribute to suppressing aggressive behaviors, their mechanisms and effectiveness can vary based on the nature of the intervention, intensity, and duration. The results highlight the enhancement of emotional regulation as a key pathway through which exercise lowers impulsive aggression. Exercise has been shown to alleviate anxiety and anger, leading to a decrease in emotional aggression. However, for instrumental aggression, which relies heavily on cognitive strategies and motivation, longer-term interventions may be necessary to achieve significant improvement. Similarly, team sports foster social connections and a sense of belonging, which further suppress aggression stemming from group conflicts. Cultural background and individual traits significantly influence the efficacy of exercise interventions. In collectivist cultures, group exercises enhance social connection and belonging, effectively mitigating aggression. Conversely, in individualistic cultures, individual exercise is more frequently viewed as a method for emotional management, helping individuals alleviate stress and anxiety. This distinction underscores the adaptability and potential of exercise interventions across various contexts. The paper also discusses the interplay between emotional regulation and cognitive control, demonstrating that they can reinforce each other in reducing aggression. Emotional regulation provides a stable psychological foundation for cognitive control, while improved cognitive control enables individuals to manage their emotions more effectively during conflicts. This bidirectional interaction, however, is subject to individual characteristics and the type of exercise performed. Furthermore, the model emphasizes how social connection and social identity recognition interact at the group level, mitigating in-group and out-group conflicts. Behavioral synchrony in group exercises enhances emotional resonance and boosts social identity recognition, which collectively contribute to lowering aggression. However, the potential "double-edged sword" effect of strong in-group identity may lead to increased out-group hostility in certain contexts. Neurobiological and physiological mechanisms are also discussed as foundational supports for exercise interventions. Key factors such as Brain-Derived Neurotrophic Factor (BDNF) levels, oxytocin release, and the functionality of the prefrontal cortex and amygdala play critical roles in emotional regulation and cognitive control. Exercise has been shown to enhance BDNF levels, improve the functional connectivity between the prefrontal cortex and amygdala, and regulate cortisol levels, all of which are integral to suppressing aggressive behavior. In summary, this research contributes to a deeper understanding of the mechanisms through which exercise interventions suppress aggressive behavior. The proposed dynamic interactive comprehensive model provides a novel theoretical framework for future studies aimed at optimizing intervention strategies. The integration of individual and group levels, alongside the consideration of cultural contexts and individual characteristics, highlights the complexity of aggression management through exercise. Future research is encouraged to validate and refine this model, utilizing advanced methodologies and exploring the effects of exercise interventions on aggressive behavior in diverse populations. Table and Figures | Reference | Related Articles | Metrics

Select

The influence of algorithmic human resource management on employee algorithmic coping behavior and job performance XI Meng, LIU Yue-Yue, LI Xin, LI Jia-Xin, SHI Jia-Zhen Advances in Psychological Science    2025, 33 (6): 948-964.   DOI: 10.3724/SP.J.1042.2025.0948 Abstract （1358）   HTML （68）    PDF （690KB）（8877）       Knowledge map    Algorithmic human resource management (HRM) is an emerging research field that combines artificial intelligence (AI) with HRM, representing a transformative shift in the field of strategic HRM and emphasizing the use of data-driven algorithms to enhance decision-making processes and optimize workforce management. While its operational benefits are widely recognized, its deeper implications for employee job performance remain underexplored, particularly in the context of employees' perceptions, trust, and behavioral adaptations to algorithmic systems. This study addresses these gaps by offering a nuanced theoretical framework that investigates the mechanisms through which algorithmic HRM influences employee job performance by examining the mediating role of employees' cognitive and emotional responses, as well as their algorithmic coping behaviors. This research builds on structuration theory to explore the duality of technology and human agency in algorithmic HRM. Specifically, it positions employees not merely as passive recipients of algorithm-driven decisions but as active agents who interpret, adapt, or resist these technologies. By integrating structuration theory's emphasis on the interplay between structural constraints and human agency, this study highlights how employees' perceptions of algorithmic transparency, fairness, and trust shape their cognitive, emotional, and behavioral responses. Furthermore, it underscores the importance of balancing algorithmic efficiency with ethical considerations to sustain employee engagement and organizational legitimacy. The innovative contributions of this study include a differentiation between the impacts of algorithmic HRM on in-role performance and extra-role performance. The study theorizes that while algorithmic precision and real-time feedback enhance task performance by providing clear metrics and actionable insights, perceptions of fairness and transparency are critical for fostering trust and encouraging extra-role behaviors. This dual focus on performance dimensions provides a more holistic understanding of algorithmic HRM's effects, addressing prior research limitations that predominantly focus on operational efficiency. The study proposes several mechanisms through which algorithmic HRM influences employee performance. First, employees' perceptions of fairness and trust in algorithmic decision-making processes act as critical mediators. Transparent algorithms enhance trust, reduce resistance, and encourage engagement, while opaque or biased algorithms can elicit skepticism and hinder performance. Second, algorithmic HRM directly improves in-role performance by providing precise, data-driven guidance and individualized feedback. In contrast, extra-role performance, such as helping behaviors, relies heavily on employees' perceptions of algorithmic fairness and the degree to which algorithms respect individual circumstances. Third, the study categorizes employees' behavioral adaptations into three types: adaptation, resistance, and manipulation. Employees who adapt to algorithmic systems are more likely to achieve high in-role performance, while those who resist may experience diminished productivity. Manipulative behaviors, such as exploiting algorithmic vulnerabilities, may yield short-term gains but often undermine long-term performance and organizational trust. The study identifies several avenues for future research to expand the understanding of algorithmic HRM. First, future research could explore the sustained impacts of algorithmic HRM on employee performance, examining how trust and engagement evolve over time and under varying organizational contexts. Second, comparative analyses of different algorithmic HRM systems (e.g., predictive vs. evaluative algorithms) could reveal their unique effects on employee cognition, emotions, and behaviors, offering insights into their strengths and limitations for in-role and extra-role performance. Investigating the moderating effects of individual characteristics (e.g., personality traits, openness to change) and cultural contexts could deepen our understanding of how employees from diverse backgrounds interact with algorithmic systems and how these differences influence the effectiveness of algorithmic HRM. Finally, future studies should examine strategies to enhance the ethical and transparent use of algorithmic HRM, including employee involvement in algorithm design and periodic reviews to mitigate bias. Such research could bridge the gap between operational efficiency and ethical governance, ensuring that algorithmic HRM aligns with organizational values and employee expectations. By linking algorithmic HRM to employee performance through the mediating effects of cognition, emotion, and behavior, this study advances theoretical and practical understandings of algorithmic HRM's role in the digital workplace. It provides a robust framework for examining the interplay between technology and human agency, highlighting the importance of fairness, trust, and adaptability in leveraging algorithmic systems for sustainable performance gains. The findings underscore the need for a balanced approach that integrates operational efficiency with ethical and human-centered practices, offering a comprehensive roadmap for organizations navigating the complexities of algorithmic HRM. Table and Figures | Reference | Related Articles | Metrics

Select

Driving mechanisms and impact effects of AI feedback-seeking behavior: A research proposal SUN Fang, LI Shaolong, LONG Lirong, LEI Xuan, ZENG Xianglin, HUANG Xiahong Advances in Psychological Science    2025, 33 (10): 1647-1662.   DOI: 10.3724/SP.J.1042.2025.1647 Abstract （1318）   HTML （186）    PDF （710KB）（4535）       Knowledge map    In the current VUCA (volatility, uncertainty, complexity, ambiguity) era, employees must proactively seek feedback to facilitate personal development and enhance their workplace competitiveness. Artificial intelligence (AI) offers new opportunities for proactive feedback-seeking, with a survey by Oracle Corporation indicating that over 50% of employees prefer seeking feedback from AI systems. However, traditional research on feedback-seeking behavior has yet to incorporate AI as a feedback source, leaving the mechanisms and consequences of employee feedback-seeking from AI largely underexplored. Moreover, emerging studies on AI feedback primarily position employees as passive feedback recipients, paying limited attention to their proactive feedback-seeking behaviors. Therefore, this research aims to bridge the gap by integrating insights from traditional feedback-seeking behavior literature with emerging studies on AI feedback, expanding the concept of feedback-seeking to include AI as a viable source, and contributing to the burgeoning field of emerging technologies and employee psychology and behavior. Specifically，Study 1 adopts a human-AI interaction perspective to examine how AI system characteristics influence employees’ feedback-seeking from AI, as well as the underlying mechanisms. Drawing on mind perception theory and trust literature, this study proposes that employees’ perceptions of AI agency and experience, along with their cognition-based and affect-based trust in AI, serve as serial mediators in the relationships between two AI system features—transparency and anthropomorphism—and employees’ feedback-seeking from AI. Furthermore, acknowledging that task characteristics are essential factors in traditional feedback-seeking research, we suggest that problem solving moderates the serial mediated relationship. Building on the core outcome of interest in the feedback literature, Study 2 seeks to explore the consequences of employees’ feedback-seeking from AI on performance improvement and its underlying mechanisms. Based on feedback process theory and AI-related literature, this study proposes that feedback-seeking from AI positively predicts employees’ performance improvement. This effect is mediated by the accuracy and specificity of AI-generated feedback information. Moreover, the type of task—objective vs. subjective—is expected to moderate this mediation process. In sum, the findings from the two studies offer several important theoretical contributions. First, this research innovatively positions artificial intelligence as a target of employees’ feedback-seeking behavior, thereby expanding the boundaries of the feedback-seeking literature. It lays a foundation for future research to explore the unique dynamics, antecedents, and consequences of seeking feedback from AI in the workplace. Second, by focusing on two critical system characteristics—transparency and anthropomorphism—this research identifies how system-level features, which have been emphasized in evaluations of AI by human users, influence employees’ feedback-seeking from AI. It reveals unique antecedents and mechanisms through which AI system features shape user behavior, deepens our understanding of the relationship between system characteristics and feedback-seeking behavior, and provides theoretical guidance for future research in AI-enabled work settings. Third, taking into account the distinctive nature of human-AI interaction, this research examines the consequences of feedback-seeking from AI on employee performance-related outcomes and uncovers the underlying mechanisms. Performance is a core outcome of interest in feedback-seeking research, yet prior studies have yielded inconsistent findings regarding the feedback-performance link. By exploring how feedback-seeking from AI influences performance improvement and through which mechanisms, this research helps clarify this relationship and broadens the theoretical boundaries of AI-related feedback-seeking research. From a practical perspective, the findings offer actionable insights for managers seeking to encourage employees to proactively seek feedback from AI systems, helping them leverage potential benefits while mitigating associated risks. It also highlights the types of work tasks where such feedback-seeking is most appropriate and effective. Overall, this research positions AI feedback-seeking as a lens through which managers can better understand the evolving interplay between emerging technologies and employee behavior, and it invites practitioners to rethink how to foster synergy between humans and intelligent systems. Table and Figures | Reference | Related Articles | Metrics

Select

Artificial theory of mind in large language models: Evidence, conceptualization, and challenges DU Chuanchen, ZHENG Yuanxia, GUO Qianqian, LIU Guoxiong Advances in Psychological Science    2025, 33 (12): 2027-2042.   DOI: 10.3724/SP.J.1042.2025.2027 Abstract （1311）   HTML （128）    PDF （587KB）（1389）       Knowledge map    In recent years, the rapid development of artificial intelligence (AI) has continuously reshaped our understanding of its capability boundaries. Evaluating the theory of mind capabilities of large language models (LLMs) has received significant attention within the research community. Recent studies suggest that LLMs can successfully complete tasks traditionally used to assess theory of mind in humans. However, controversial questions remain: Do LLMs possess theory of mind? What are the essential differences between artificial theory of mind and human theory of mind? Therefore, this systematic review synthesizes the performance of LLMs on theory of mind tasks, reveals essential differences in the internal processes between human theory of mind and artificial theory of mind, refines the conceptual definition of artificial theory of mind, and outlines key challenges in this field. Specifically, we systematically synthesize research on artificial theory of mind from the objects of evaluation and the characteristics of tasks. Following the developmental sequence of core theory of mind sub-abilities in humans, we evaluate the task performance of GPT-4. Results indicate that GPT-4 can consistently pass false-belief tasks and various higher-order theory of mind tasks, suggesting a simulation of human-like theory of mind performance. Nevertheless, recognizing that behavioral task accuracy may be insufficient to reflect true capability, we specifically examine cases in which model performance is limited. Our analysis identifies critical limiting factors including external factors (e.g., limitations in test items, prompt design and human baselines) and intrinsic limitations of LLMs (e.g., hallucinations, hyperconservatism, commonsense errors, heuristics or spurious correlations, and spurious causal inference). This suggests that performance fluctuations may not stem from a lack of artificial theory of mind. Consequently, by integrating GPT-4's high accuracy with its intrinsic and extrinsic factors limiting its performance, we demonstrate that GPT-4 has developed artificial theory of mind that is similar in performance to human theory of mind. The advanced capabilities of LLMs underscore the importance of studying the internal processes of theory of mind. Multiple lines of evidence suggest fundamental differences in the internal processes between human theory of mind and artificial theory of mind. To fully distinguish them, we conduct a comparison of the neural foundations and developmental factors supporting both. We highlight that their neural mechanisms exhibit distinct complexity across multiple dimensions, and the factors influencing theory of mind acquisition and development differ between humans and LLMs. This reveals the root cause of their internal process differences. Based on this, we define artificial theory of mind as: the simulation of human theory of mind performance exhibited by LLMs during text generation, achieved through recognizing and matching statistical patterns in the text after receiving a theory of mind task prompts. However, as an emerging field, existing research faces three main challenges: lack of standardized evaluation protocols, unclear mechanisms of artificial theory of mind, and theory of mind alignment issues. Regarding evaluation, issues in experimental design, prompt design, and scoring methods can lead to divergent results. Regarding mechanisms, while internal processes are key to distinguishing between theory of mind and artificial theory of mind, current research has neither fully elucidated these processes nor sufficiently addressed​ how humans attribute mental states to LLMs. Regarding alignment, LLMs simulate human theory of mind performance without achieving genuine theory of mind reasoning. Based on this, we propose corresponding research directions. In conclusion, this study reveals that LLMs possess an artificial theory of mind that is similar in performance to human theory of mind but different in internal processes, refines its conceptual definition, and clarifies key field challenges. It identifies critical issues and delineates future research directions, offering a foundation for leveraging artificial theory of mind research to advance our understanding of human theory of mind emergence and internal processes. Table and Figures | Reference | Related Articles | Metrics

Select

The psychotherapeutic mechanisms and neural basis of Eastern mindfulness WU Kai Advances in Psychological Science    2026, 34 (2): 331-347.   DOI: 10.3724/SP.J.1042.2026.0331 Abstract （1278）   HTML （56）    PDF （792KB）（1019）       Knowledge map    This paper introduces an innovative “therapy-neural-culture continuum” framework that reconceptualizes mindfulness as a holistic psychological intervention, elucidating the unique psychotherapeutic mechanisms and neural foundations of Eastern mindfulness. Rooted in Chinese Mahāyāna (Han) Buddhism and integrated with Confucian and Daoist philosophies, Eastern mindfulness addresses the limitations of Western mindfulness models, which have often undergone decontextualization and ethical omission, leading to conceptual reductionism and potential adverse effects. By restoring the integral Three Trainings—śīla (morality), samādhi (concentration), and prajñā (wisdom)—as a unified system of mental cultivation, this approach transcends individualistic and instrumental adaptations, providing a culturally grounded pathway for deep psychological transformation and enhanced intervention safety. A central innovation lies in the psychological reinterpretation of Eastern mindfulness’s theoretical core, distilled into four interrelated pillars: emptiness (śūnyatā), Buddha-nature (tathāgatagarbha), bodhisattva practice, and non-duality. Emptiness is reframed as cognitive de-reification, enabling practitioners to perceive thoughts and self-concepts as transient processes rather than fixed entities. This aligns with the defusion process in acceptance and commitment therapy (ACT), yet surpasses cognitive behavioral therapy’s (CBT) reappraisal techniques by fundamentally challenging the reality of mental constructs. Buddha-nature reframes healing as the realization of inherent potential rather than the correction of deficits, paralleling posttraumatic growth (PTG) theory in which trauma catalyzes resilience and wisdom beyond mere recovery. Bodhisattva practice transforms motivation from self-centered relief to altruistic fulfillment, promoting prosocial behavior through compassion training—supported by evidence of enhanced empathy, gratitude, and reduced bias. Non-duality fosters psychological flexibility, integrating opposing experiences such as pleasure and pain, enabling value-driven living amid distress, as demonstrated in studies of emotional adaptability. Collectively, these pillars define Eastern mindfulness as a wisdom-oriented psychotherapeutic model that unites ethical, cognitive, and relational dimensions. The practice architecture is reconstructed through the Three Trainings as an interdependent psychological model. Morality (śīla) functions as ethical self-regulation, reducing inner conflict and strengthening psychological immunity through disciplined conduct, establishing a secure foundation for deeper contemplative work. Concentration (samādhi) enhances attentional stability and emotional regulation via focused-attention and loving-kindness meditation, bridging external behavioral harmony with internal cognitive clarity. Wisdom (prajñā) culminates in metacognitive insight, paralleling open monitoring meditation, deconstructing the illusion of self for profound transformation. Unlike Theravāda’s emphasis on individual liberation, Mahāyāna’s prajñā integrates altruistic insight, emphasizing compassionate awakening. This spiral progression ensures comprehensiveness, as each training reciprocally reinforces the others, forming a dynamic system of behavioral ethics, cognitive clarity, and transcendental understanding. The study further outlines these mappings through conceptual modeling, highlighting how Eastern mindfulness fuses Buddhist ethics, Confucian moral harmony, and Daoist natural balance into a relational, holistic approach that surpasses Western models focused solely on attention and awareness. Clinically, although direct randomized controlled trials (RCTs) on the integrated Eastern framework are limited, aggregated evidence from related interventions supports its efficacy. Compassion and loving-kindness meditation reduce anxiety, enhance positive affect, and improve social cognition. Body-mind practices such as Qigong and Baduanjin alleviate depression, stress, and inflammation by upregulating serotonin and brain-derived neurotrophic factor (BDNF), and by modulating autonomic nervous system activity. Ethical and compassion-based practices reduce compassion fatigue and strengthen prosocial connectedness, while wisdom-oriented practices foster non-dual awareness and self-transcendence, mitigating self-centered cognition. At the neural-physiological level, the model innovatively delineates multi-pathway integrations. Morality engages compassion-prosocial circuits, activating the dorsomedial prefrontal cortex (dmPFC) and temporoparietal junction (TPJ) for empathy while downregulating the amygdala to achieve emotional equilibrium, with gamma synchrony marking stable trait changes. Concentration modulates default mode-executive control network (DMN-ECN) dynamics, reducing rumination via posterior cingulate cortex (PCC) deactivation and enhancing executive regulation through anterior cingulate (ACC) and prefrontal (PFC) engagement. Wisdom activates insight-deconstruction pathways, diminishing self-referential processing in the DMN and amplifying gamma/theta oscillations linked to cognitive flexibility. Systemic effects manifest through the neuroendocrine-immune axis, where HPA-axis downregulation lowers cortisol, enhanced vagal tone increases HRV and parasympathetic dominance, and anti-inflammatory shifts (e.g., interleukin-6 reduction) foster resilience. Cultural neuroscience further clarifies that repetitive engagement in these practices shapes “culturally patterned neural activity”, accounting for Eastern mindfulness’s superior safety, ethical depth, and integrative efficacy. Future work should focus on empirical verification, protocol standardization, and cross-cultural adaptation. The paper calls for high-quality RCTs, neuroimaging validation, and indigenous psychometric development to establish Eastern mindfulness as a scientifically robust Chinese psychotherapy paradigm. Ultimately, this therapy-neural-culture continuum not only restores mindfulness’s Buddhist essence but also positions it as a bridge linking therapy, neuroscience, and culture, offering globally relevant insights for ethical, transformative, and culturally attuned mental health interventions, and expanding the scientific boundaries of mindfulness research. Table and Figures | Reference | Related Articles | Metrics