The rapid advancement of Large Language Models (LLMs) is catalyzing a profound transformation in psychological research. This review advances a dual-view framework for critically assessing LLMs: as powerful simulators of human participants and as novel non-human agents for scientific inquiry. As simulators, LLMs demonstrate considerable potential in replicating human responses. Empirical studies have shown that models can mimic human cognitive biases in decision-making tasks, achieve performance on the Theory of Mind (ToM) tasks comparable to that of young children, and successfully replicate a significant portion of main and interaction effects from classic psychology experiments. However, this paper moves beyond cataloging these successes to deconstruct the simulator concept itself, revealing two fundamental, principle-level limitations. First, the model's core compression principle, while effective at capturing group averages, inherently struggles to replicate the internal structural associations of individual differences. Second, a profound asymmetry exists between the model's training phase (knowledge acquisition) and its alignment phase (strategic expression), which fundamentally determines that any simulation is a strategically optimized expression rather than a faithful representation of the model's internal knowledge state. Consequently, their inherent limitations and systematic biases - such as reflecting a predominantly Western, Educated, Industrialized, Rich, and Democratic (WEIRD) perspective - are not merely data-level flaws, but deep-seated mechanistic properties. This transforms them into valuable data for investigating the mechanisms of non-human intelligence. This critical perspective aligns with recent scholarly calls to systematically analyze the theoretical fallacies, such as anthropomorphism and identity essentialization, that arise from uncritically substituting LLMs for human participants.
Within this theoretical framework, the paper systematically organizes the current methodologies into two primary technical paths, distinguished by their depth of intervention. The first is the path of prompt engineering, which guides the model's output without altering its internal parameters. This path encompasses techniques for individual-agent simulation, where methods like simple prompts are used to assign specific personas to the model, as well as techniques for group-level simulation, where in-context learning allows the model to predict aggregate public opinion by generalizing from a few examples. The second, more intensive path is that of model fine-tuning, which directly modifies the model's parameters by retraining it on large-scale, individual-level datasets to enhance its predictive accuracy for specific populations. To rigorously assess the outputs from these methods, the review scrutinizes two key evaluation frameworks. The first, algorithmic fidelity, evaluates the degree to which an LLM can mirror the complex relationships between thoughts, attitudes, and socio-cultural backgrounds within specific human subpopulations. The second, the Turing Experiment, assesses an LLM's ability to replicate the behavioral outcomes of human subjects in classic social science experiments, focusing on functional equivalence in contexts like the Ultimatum Game or Milgram's obedience studies, while also critiquing the fallacy of “perfect alignment” as a goal.
Finally, this review provides crucial suggestions for leveraging LLMs in psychological research. It strongly emphasizes that LLMs cannot replace human participants due to fundamental limitations, such as the lack of embodied cognition, subjective experience, and genuine understanding. This is particularly evident in their inability to natively generate non-textual data crucial for many fields, such as reaction times in cognitive psychology or physiological indicators in neuroscience. Instead, the paper argues that future research should proceed along dual paths: (1) Deepening the role of LLMs as a tool and (2) Pioneering the study of LLMs as a subject. As a tool, LLMs can serve as invaluable simulators for hypothesis generation, pre-testing study materials, and conducting exploratory research on sensitive topics where human participation poses ethical risks. As a subject, researchers can investigate the LLM itself as a novel agent. For instance, they can systematical map the ideological spectrums of different models, analyze their "hallucinations" as a form of computational creativity, or design "implicit association tests" to uncover their deep-seated biases. This dual-path approach ensures that this powerful technology is leveraged rigorously and responsibly for knowledge discovery. Ultimately, it highlights the irreplaceable value of human researchers in proposing profound questions, interpreting complexity, and defending scientific integrity.
Recent advances in artificial intelligence (AI), particularly large language models (LLMs), have profoundly reshaped both technological landscapes and scientific inquiry in psychology. This paper systematically examines the bidirectional synergy between psychometrics and AI, arguing that their integration is not merely instrumental but foundational for the emergence of a new measurement paradigm in the era of human-AI coexistence.
On one hand, psychometrics provides essential theoretical frameworks and methodological rigor to evaluate AI systems. By adapting established psychological constructs—such as intelligence, creativity, personality, theory of mind, and moral reasoning—researchers can assess AI’s “psychological” profiles, benchmark its capabilities against human standards, and uncover its developmental trajectories and limitations. Current approaches include the direct application of human psychometric instruments (e.g., WAIS, Big Five inventories) and large-scale AI-specific benchmarks (e.g., MMLU, BIG-bench). Empirical studies show that models like GPT-4 already match or exceed human performance in domains such as verbal reasoning and creativity fluency, yet lag in visual-spatial tasks and emotional depth. However, these methods often suffer from anthropomorphic bias, prompt sensitivity, and a lack of grounding in psychological theory, leading to questionable validity. Moreover, most evaluations rely on classical test theory, yielding ordinal rankings that hinder fine-grained, cross-model comparisons on a common metric. To address this, we advocate for integrating modern psychometric models—such as item response theory and cognitive diagnosis models—to enable equated, interpretable, and diagnostic assessments of AI capabilities. Beyond trait measurement, psychometrics also enables the systematic study of (1) human cognition, emotion, and attitudes toward AI; (2) AI’s impact on human psychological development in education, mental health, and socialization; and (3) interdependent dynamics in human-AI collaboration, including role allocation, trust calibration, and interaction patterns.
On the other hand, AI is revolutionizing psychometrics itself. LLMs facilitate automated item generation, significantly reducing development costs and human bias while enabling dynamic, context-sensitive assessments. AI also supports implicit and multimodal measurement through the analysis of natural language, facial expressions, voice, and behavioral logs, moving beyond traditional self-report questionnaires. Furthermore, deep learning enables the unsupervised extraction of latent psychological dimensions from real-world data (e.g., social media), potentially refining or even redefining psychological constructs. In scoring and interpretation, AI systems can provide reliable, scalable, and diagnostic feedback on open-ended responses, while predictive modeling allows for early risk detection and personalized interventions. For instance, transformer-based models like CLIP and Flamingo enable cross-modal integration of text, image, and audio, while graph neural networks model complex problem-solving trajectories. Nevertheless, these advances raise critical concerns about algorithmic “black boxes,” data bias, cross-cultural fairness, and the ethical use of sensitive behavioral data.
The paper identifies key challenges in both directions. For psychometrics-to-AI, issues include the risk of uncritical anthropomorphism, unstable AI responses due to prompt and parameter sensitivity, cultural bias in benchmarks, and the lack of fine-grained diagnostic feedback in current evaluations. For AI-to-psychometrics, concerns center on transparency, the validity of AI-generated content, data privacy, and the scarcity of interdisciplinary expertise. To address these, we propose five strategies: (1) developing AI-specific psychometric paradigms that account for AI’s “data-algorithm-model” nature and incorporate functional traits (e.g., reasoning stability, cross-context adaptability); (2) creating stability metrics, such as output consistency indices, to quantify AI’s trait volatility; (3) embedding core psychometric principles—reliability, validity, fairness—into AI systems from the design stage; (4) training domain-specific “AI psychometricians” that integrate psychological theory with computational methods; and (5) establishing ethical guidelines for data collection, use, and synthetic data generation.
Ultimately, we envision a co-evolutionary future in which psychometrics and AI mutually inform each other: psychometrics offers interpretability, standardization, and ethical grounding, while AI contributes scalability, adaptivity, and multimodal integration. Their deep integration may give rise to intelligent psychometrics—a new discipline that merges theory-driven and data-driven approaches to understand both human and artificial minds in an increasingly intertwined world. This synergy not only advances scientific understanding but also ensures that AI development remains human-centered, scientifically sound, and ethically responsible.
Mental health issues not only affect individuals’ quality of life and social functioning, but also pose broader risks to social well-being. Cognitive Behavioral Therapy (CBT), a widely used approach in mental healthcare, focuses on cognitive restructuring—modifying emotional and behavioral responses by identifying, challenging, and transforming negative or distorted automatic thoughts. Although CBT has demonstrated substantial effectiveness in clinical settings, its widespread implementation remains constrained by resource scarcity, high costs, and limited therapist availability. With the rapid advancement of artificial intelligence (AI), AI-powered CBT-style support systems offer a promising wang to deliver scalable, low-cost psychological support.
This study proposes a novel CBT-CoT (Cognitive Behavioral Therapy with Chain-of-Thought reasoning) framework, which combines CBT’s structured psychological reasoning process with a multi-agent architecture driven by large language models (LLMs). The framework is organized into two key stages—cognitive assessment and cognitive restructuring—and implemented through four collaborative and functionally distinct specialized agents.
In the cognitive assessment stage, the Automatic Thought Extraction Agent structurally decomposes the user’s message into a standardized three-column record comprising the precipitating event, the associated automatic thoughts, and the resulting emotional and behavioral responses. The Automatic Thought Evaluation Agent then analyzes these automatic thoughts by examining whether they are facts or interpretations, their belief intensity, and the types of cognitive distortions, and determines whether the conversation should proceed to the next stage of structured reasoning.
If restructuring is required, the CoT Reasoning Agent matches appropriate cognitive restructuring techniques and generates alternative rational thoughts by employing reasoning pathways such as hypothesis verification and evidence analysis. This agent outputs a structured five-column record aligned with standard CBT practice. Finally, the Response Generation Agent integrates the five-column record and the selected restructuring strategy to produce a personalized, empathetic, and cognitively supportive response tailored to the user’s concern.
Building on the methodology of PsyQA, we constructed a new large-scale Chinese psychological question-answering dataset comprising over 20,000 question-answer pairs across nine key domains, including interpersonal relationships, emotional adjustment, academic stress, personal growth, and others. The CBT-CoT system adopts a modular architecture, in which the four agents cooperate to complete a full-cycle CBT process—from the extraction and evaluation of automatic thoughts to the generation of supportive language based on structured reasoning.
To identify the most suitable base model for CBT-CoT, we evaluated three state-of-the-art Chinese LLMs: DeepSeek-Chat, Qwen-Plus, and Doubao-1.5-pro. DeepSeek-Chat consistently outperformed the other models in both reasoning quality and response fluency, and was selected as the base model for all four agents in the CBT-CoT framework.
Evaluation was conducted using both automated and manual methods. An ensemble of three large language models served as automated evaluators, while trained annotators performed human evaluations on a stratified sample. The results showed that CBT-CoT significantly outperformed baseline methods across six dimensions of psychological support: empathy expression, cognitive restructuring effectiveness, depth of reasoning, appropriateness of strategy, logical coherence, and degree of personalization. The total scores of human evaluation and LLM-based automatic scoring across the six dimensions were 17.18 and 17.48, respectively, with an absolute difference of only 0.30 points. In comparison with human-written responses (total score: 7.97), the CBT-CoT system (total score: 17.48) consistently generated more structured, coherent, and therapeutically oriented replies.
In summary, the CBT-CoT framework operationalizes the core cognitive restructuring flow commonly emphasized in CBT practice through multi-agent collaboration. By dividing the process into cognitive assessment and restructuring stages, and by explicitly modeling the chain-of-thought reasoning path, the system provides structured, explainable, and supportive responses. The framework demonstrates robust generalizability across diverse psychological topics and user intents. Future research may further optimize the CBT-CoT framework to improve its accuracy and adaptability in addressing complex psychological issues, while ensuring the integration of ethical safeguards into real-world deployment scenarios.
In daily life, individuals exert efforts in pursuit of diverse rewarding objectives. Such efforts can be categorized as either self-directed (pro-self efforts) or others/society-directed (prosocial efforts). Prosocial behaviors, which entail expending efforts for the welfare of others, hold a significant position within the realm of human moral conduct and are of crucial importance for human well-being and the sound progress of society. However, the majority of prior research has primarily focused on the impact of pro-self efforts on reward processing, with less attention given to how prosocial efforts influence this process. Furthermore, the manner in which the reward value associated with prosocial efforts dynamically evolves over time remains unclear. The present study employed the effort-door task, integrated with event-related potential (ERP) technology, to examine the impact of pro-self and prosocial effort on the habituation of reward processing at different stages (i.e., “anticipation” and “consumption”).
The study included 42 participants and adopted a 2 (effort type: pro-self vs. prosocial) × 2 (effort level: high vs. low) within-subjects design. Pro-self effort was defined as the effort exerted to win money for oneself, whereas prosocial effort referred to the effort exerted to win money for a charity. Effort level was manipulated via the required number of key presses. To examine the neural dynamics of reward processing, we employed Cue-P3 and reward positivity (RewP) as neural correlates of reward anticipation and consumption, respectively. Each trial began with a 1000 ms fixation cross, followed by a lock icon and the required number of key presses. Participants used the pinky finger of their non-dominant hand to press the "H" key until the required effort was completed, at which point the lock icon turned into an unlocked symbol. Then, two identical doors appeared on the screen, from which participants selected one to receive feedback. Feedback was presented as either a green upward arrow (indicating a gain of 1 yuan) or a red downward arrow (indicating a loss of.5 yuan), each with a 50% probability.
Considering the nested nature of the data, we used 4 multilevel linear models (MLM) to examine the effects of effort type and effort level on reward anticipation and consumption habituation. A multilevel linear model for Cue-P3 revealed a significant main effect of effort level (B = -.66, SE=.32, t(574) = -2.06, p <.05), with higher amplitudes under high effort. No other main effects or interactions were significant. For RewP, a significant three-way interaction emerged (B = -.18, SE =.09, t(2613) = -2.05, p <.05). Specifically, for pro-self effort, the main effect of effort level (B = 1.67, SE =.70, t(301) = 2.39, p <.05) and its interaction with trial number (B = -.20, SE=.06, t(1262) = -3.25, p <.01) were significant. Simple slope analysis showed no trial effect under high pro-self effort(B=.06, SE =.05, t(103) = 1.15, p >.05), but a significant decrease under low pro-self effort(B= -.14, SE =.05, t(95.3) = -2.90, p <.01). For prosocial effort, only a significant main effect of trial number was found (B= -.11, SE =.04, t(980) = -2.65, p <.01), with no other significant effects (the main effect of effort level: B=.52, SE =.68, t(1277) =.77, p >.05; the interaction between effort level and trial: B = -.02, SE =.06, t(1304)= -.33, p >.05).
These findings demonstrate that both pro-self and prosocial efforts enhance reward anticipation, but do not influence its habituation process. Moreover, pro-self effort slows down the habituation of pleasure derived from rewards in the reward consumption phase, whereas prosocial effort does not affect this process. This indicates a dissociation between the aftereffects of pro-self and prosocial efforts on reward consumption habituation but not on reward anticipation habituation. The research expands our understanding of the temporal dynamics of reward value driven by different motivations. It also holds significant implications for understanding how humans perceive and adapt to rewards under various motivational drives..
Understanding the differences in the sequential decision-making between Large Language Models (LLMs) and humans is crucial for optimizing human-AI collaboration. This study employed the Iowa Gambling Task (IGT), combined with Value plus Sequential Exploration (VSE) cognitive modeling, to systematically compare behavioral performance between humans and LLMs. We contrasted a large human sample with three representative LLMs under baseline and Chain-of-Thought (CoT) conditions. We hypothesized that the VSE model could effectively characterize the exploration and exploitation parameters in samples including both humans and LLMs.
The study utilized a clinical version of the IGT with dynamic reward/loss schedules. Data were collected from 1763 Chinese undergraduate students via a standard computerized procedure. LLM data were collected via the official APIs of three models (DeepSeek-V3, GLM-4-Flash, GLM-Zero-Preview), simulating 48 participants per model for both baseline and CoT conditions. To prevent LLMs from leveraging pre-existing knowledge about the IGT task acquired during their training, the standard deck labels (A, B, C, D) were replaced with random three-letter strings as anonymized labels. Furthermore, the assignment of the underlying deck characteristics (i.e., their reward and punishment schedules) to the anonymized labels was counterbalanced across participants using a Latin square design. LLM interactions involved system prompts delivering instructions and user prompts requesting choices. They incorporated dialogue history and feedback on gains/losses from the previous trial. An error-correction mechanism ensured valid responses. Cognitive modeling compared the VSE model against four alternative models using Variational Bayesian Analysis for parameter estimation and model fitting with non-informative priors. A parallel analysis approach was adopted. We conducted comparisons between the LLMs and (1) the entire human sample, and (2) the top-performing (based on final net gain) 48 “expert” human participants, matched to the LLM sample size.
The behavioral results indicated that certain LLM conditions (baseline GLM-4-Flash and GLM-Zero-Preview) failed to follow IGT task instructions. They exhibited repetitive non-adaptive patterns and were thus excluded from further analysis. LLMs (Deepseek-V3 and GLM-Zero-Preview) achieved significantly higher final net gains compared to the overall human sample, but their performance showed no significant difference from that of the “expert” human sample. A comprehensive investigation involving model comparison analyses robustly identified the VSE model as the superior framework for explaining the choice sequences of both humans and LLMs across two parallel analyses. This conclusion was evidenced by better-fit indices (BIC, AIC, Free Energy) and higher predictive accuracy for choices. Parameter estimates from the VSE model revealed several key differences between humans and LLMs. Regarding value sensitivity (accuracy in perceiving reward/loss utility), LLMs showed greater sensitivity than the general human population, although comparable to that of expert humans. Conversely, human participants always showed a superior ability to integrate historical value information over time, reflected in higher inverse decay rate parameters compared to LLMs. Differences in exploration-related parameters (learning rate and gain) were also observed, indicating distinct exploration tendencies between the groups and across two parallel analyses. Finally, LLMs consistently demonstrated significantly higher decision stability (consistency) than human samples.
In conclusion, this research demonstrates the utility of combining the IGT, cognitive modeling, and a parallel analysis framework to dissect the nuances of sequential decision-making in both humans and LLMs. Although high-performing LLMs can achieve outcomes similar to those of human experts in the IGT, their underlying cognitive parameters differ substantially. Key distinctions include the LLMs’ heightened value sensitivity (similar to human experts) and greater strategic consistency, contrasted with humans’ superior integration of past outcomes, which is potentially linked to affective systems like somatic markers absent in LLMs. The VSE model provides a unified computational lens to examine these exploration-exploitation dynamics. These findings highlight the unique cognitive profile of LLMs. These findings offer insights for developing targeted training to enhance decision-making. They also inform the design of more effective human-LLM collaborative systems that capitalize on the distinct strengths of each agent, while acknowledging potential LLM limitations in less structured environments.
Current research on trust has primarily examined the independent effects of facial trustworthiness and social value orientation (SVO), without systematically investigating their joint influence on both one-shot trust and trust learning, and their underlying mechanisms. Furthermore, existing studies employing computational models to explore the dynamic processes of trust learning have not adequately incorporated SVO as a key factor of individual difference. Grounded in the person-context tnteraction theory and the dyadic model of trust development, facial trustworthiness as an external cue and SVO as an intrinsic trait should jointly shape trust behavior. Based on classical trust theories, such as the social norm theory and moral norm theory, reciprocity expectation may serve as the underlying psychological mechanism through which both factors influence trust. Similarly, drawing on the neuropsychological economic model of trust, the dynamic interaction between initial facial trustworthiness and feedback probability in trust learning is likely to vary as a function of an individual’s SVO, and such differences should be reflected in the computational modeling of trust learning. To address these research gaps and test these theoretical hypotheses, the present study combined the trust game paradigm with reinforcement learning modeling across two experiments to systematically examine the joint effects of facial trustworthiness and SVO at different stages of trust development (one-shot trust and trust learning) and to quantify key computational mechanisms underlying trust learning.
Experiment 1 employed a one-shot trust game to examine whether facial trustworthiness and SVO interactively influence one-shot trust and whether reciprocity expectation mediates their effects. A 2 (facial trustworthiness: high vs. low) × 2 (SVO: prosocial vs. proself) mixed design was adopted, with facial trustworthiness as a within-subjects factor and SVO as a between-subjects factor. The dependent variables were the amount invested and reciprocity expectation. The results showed that both facial trustworthiness and SVO significantly and positively predicted one-shot trust, but their interaction was not significant. Reciprocity expectation mediated the effects of both factors on trust.
Experiment 2 used a repeated trust game to examine whether the dynamic interaction between initial facial trustworthiness and feedback probability during trust learning is moderated by SVO. A 2 (SVO: prosocial vs. proself) × 2 (initial facial trustworthiness: low vs. high) × 2 (feedback probability: high vs. low) mixed design was implemented, with the first two factors as within-subjects variables forming consistent and inconsistent conditions, and SVO as a between-subjects variable. Twelve candidate reinforcement learning models were constructed and compared to identify the best-fitting model. Statistical analyses of model parameters were conducted to elucidate the specific mechanisms by which SVO influences trust learning. The results revealed that SVO significantly moderated the interactive effects of initial facial trustworthiness and feedback probability: proself individuals trusted low-trustworthiness trustees more under high feedback probability but trusted high-trustworthiness trustees more under low feedback probability. In contrast, prosocial individuals consistently trusted high-trustworthiness trustees more regardless of feedback probability. Model comparisons showed that Model 4c, which incorporated three key mechanisms identified in prior research, provided the best fit. However, SVO had no significant effect on learning rates, the facial trustworthiness influence coefficient, or the decay rate.
The findings demonstrate that facial trustworthiness and SVO exert independent effects on one-shot trust but exhibit a joint influence in trust learning, where the dynamic interaction between initial facial trustworthiness and feedback probability is moderated by SVO. This suggests that their synergistic effect emerges specifically in dynamic interactive contexts. The study verifies that reciprocity expectation serves as a common psychological mechanism through which both factors influence trust, providing empirical support for classical trust theories. The reinforcement learning model successfully captured key computational characteristics of trust learning from previous studies. Individuals adjusted learning rates based on the consistency between initial facial trustworthiness and feedback probability. Initial facial trustworthiness exerted a persistent influence on subsequent learning. Also, the impact of prediction errors from behavioral feedback on value expectations diminished over time. This study incorporated SVO into a computational modeling framework for the first time. It also revealed that trust learning may involve additional computational processes beyond the known mechanisms. SVO might influence learning through these processes. This warrants further investigation in future research.
Approximate calculation—the ability to produce rapid, coarse numerical estimates—is a fundamental component of mathematical reasoning and reliably predicts later achievement in the field. The operational momentum effect (OME) denotes a systematic bias in approximate arithmetic whereby individuals tend to overestimate outcomes of addition and underestimate the outcomes of subtraction. The mechanism that gives rise to this effect remains unresolved. The heuristic account proposes that OME reflects a cognitive shortcut: when adding, people are biased toward amounts larger than the starting value, whereas when subtracting, they are biased toward amounts smaller than the starting value. In contrast, the attentional shift account attributes OME to a spatial displacement of attention along the mental number line (MNL): addition shifts attention to the right, while subtraction shifts it to the left. This results in overestimation for addition and underestimation for subtraction.
Previous studies have examined the behavioral origins of the OME, providing substantial support for competing accounts. However, the neural basis of the effect remains poorly understood. Complementary neuroimaging methods can address this gap: event-related potentials (ERPs) provide millisecond temporal resolution for tracking attentional shifts during approximate calculation, whereas functional magnetic resonance imaging (fMRI) offers precise spatial localization. Representational similarity analysis (RSA) furthermore enables comparison of ERP and fMRI by projecting their activity patterns into a common representational space. In the present study, we combined ERP, fMRI, and RSA to investigate behavioral performance and neural activity during addition and subtraction across directional conditions (less, neutral, more). For this purpose, sixty healthy young adults performed an arithmetic verification task, judging whether each presented addition or subtraction outcome (categorized as smaller, equal, or larger than the correct result) was correct. ERP and fMRI data were then mapped into a shared representational space, and the similarity of the neural representations at successive time points was computed. This multimodal approach aimed to elucidate the neural mechanisms underlying the OME and to determine whether the effect reflects an attentional shift along the MNL.
The results indicated that spatial attention engages approximate calculation, with larger outcomes in addition and smaller outcomes in subtraction attracting more spatial attention. Specifically, the behavioral results demonstrated that participants exhibited a tendency to accept neutral or larger outcomes in addition and neutral or smaller outcomes in subtraction, thereby indicating the presence of the OME. Furthermore, the ERP results demonstrated a significant interaction between operation and direction on the P3b component (250~350 ms), which is associated with attentional allocation. Specifically, larger outcomes in addition and smaller outcomes in subtraction elicited a larger P3b wave. The fMRI results demonstrated that larger outcomes in addition and smaller outcomes in subtraction significantly activated the right intraparietal sulcus (IPS), superior parietal lobule (SPL), and precuneus, which is related to spatial attentional shift. Furthermore, the results of the functional connectivity study indicated that the right precuneus exhibited stronger functional connectivity with the middle temporal gyrus, inferior occipital gyrus, and fusiform gyrus in the larger outcomes of addition, in comparison to the smaller outcomes. This finding suggests that the former outcomes may be allocated more attentional resources. Furthermore, the RSA results indicated that the neural representation of the OME remained consistent across various data modalities, thereby underscoring the congruence between ERP and fMRI data. This observation also reflected a transition from early visual processing to more complex cognitive processing.
Taken together, these findings indicate that the OME is a result of a spatial shift of attention along a mental number line, substantiating the attentional shift account. From a neuroscientific perspective, these results elucidate the formation mechanisms of the approximate calculation bias phenomenon and unveil the intrinsic connection between numerical cognition and spatial attention. Such insights provide a scientific foundation for efficient mathematical learning and cognitive interventions.
The influence of sense of control on loss-driven attentional capture remains a topic of debate, and the underlying regulatory mechanisms are still not fully understood. While some studies have shown that loss-related stimuli can automatically capture attention, others have failed to replicate these findings. Investigating whether and how loss-related information captures attention is essential for understanding the cognitive foundations of adaptive human behavior. Sense of control influences both attentional selection and the mitigation of loss threats. Therefore, it is likely a significant factor in modulating attentional capture related to losses.
This study utilized eye-tracking technology and applied the Hierarchical Drift Diffusion Model (HDDM) to computationally model behavioral data. The aim was to investigate the cognitive and computational mechanisms by which choice-based and outcome-based control influence loss-driven attentional capture. In Experiment 1, choice-based control was manipulated by altering the causal relationship between keypress responses and the stopping of a roulette wheel. During the associative learning phase, participants established associations between different colors and outcomes of high loss, low loss, or no loss through a roulette game. In the following test phase, the additional singleton paradigm was employed to assess the attentional capture effect of loss-associated distractors. The two key variables manipulated in Experiment 1 were the degree of control and the level of loss. In the high control condition, the roulette wheel stopped immediately on the loss score selected by the participant upon keypress. In the low-control condition, it continued spinning for a period before forcibly stopping at a computer-selected loss value. The available loss options were “-50 points,” “-5 points,” and “0 points,” corresponding to high-loss, low-loss, and no-loss outcomes, respectively. Behavioral results indicated slower search response times in the high-control condition compared to the low-control condition, suggesting a stronger attentional capture effect. Eye-tracking data showed larger pupil diameters, shorter first-fixation latencies, and longer fixation durations in the high-control condition, indicating heightened arousal levels, faster early attentional orienting, and enhanced late attentional maintenance. HDDM results revealed a lower drift rate (v) and longer non-decision time (t) in the high-control condition, reflecting a slower evidence accumulation speed during decision-making and extended stimulus encoding and response organization outside the decision process.
In Experiment 2, outcome-based control was manipulated by varying the alignment between keypress outcomes and expected losses, while ensuring a causal relation between the keypress responses and the stopping of the roulette wheel. The proportion of trials in which the outcome matched expectations was set at 100%, 75%, or 50% to correspond to high, medium, and low control, respectively. The total loss decreased as the degree of control increased. The results indicated that search accuracy was lowest and fixation durations on the target were shortest in the high-control condition compared to the low-control condition, suggesting a trade-off between processing precision and speed. HDDM results indicated that reaction times were primarily constrained by the non-decision time (t) parameter, which was not influenced by the degree of control.
These findings suggest that sense of control modulates loss-driven attentional capture through a dual-path mechanism: (1) Choice-based control enhances attentional capture by accelerating early attentional orienting and prolonging late attentional maintenance; and (2) Outcome-based control influences attentional selection by balancing processing speed and precision during late attentional maintenance. This framework sheds light on the dynamic processes through which sense of control regulates loss-driven attentional capture. Future research could further investigate its biological basis using cognitive neuroscience techniques, combined with computational neural modeling to examine predictive pathways from neural activity to behavioral responses. Additionally, studies could explore the interaction between choice-based and outcome-based control in influencing loss-driven attentional capture and compare their relative contributions to this modulation..
Although human experience unfolds continuously, memory organizes this flow into discrete, meaningful events through a process called event segmentation. Event boundaries, the points at which segments transition, are often triggered by shifts in context, goals, or perceptual information. While crucial for organizing experience, event boundaries frequently impair memory for the temporal order of information encountered across them, a phenomenon termed the event boundary effect. This effect is robust, observed across various paradigms and timescales, suggesting that boundaries disrupt the encoding or retrieval of temporal relationships between items belonging to different events. However, empirical findings regarding boundary effects have been inconsistent. While numerous studies confirm the detrimental impact of boundaries on temporal memory, some research suggests that under specific conditions, such as high unitization encoding, this impairment can be substantially reduced or even eliminated. The inconsistency of these findings and the insufficient exploration of regulatory conditions in existing studies highlight the need to systematically examine the malleability of event boundary effects and identify the key factors that modulate their impact on temporal order memory..
Following an a priori power analysis, fifty undergraduate participants (27 male; Mage = 19.68 years, SD = 1.56 years) were recruited for Experiment 1. Participants were randomly assigned to context-present or context-absent retrieval conditions. During encoding, they viewed sequences of object-scene background pairs designed to produce three levels of bottom-up unitization: high (semantically coherent backgrounds), medium (moderately coherent), and low (incoherent). Each sequence contained an event boundary defined by a background context change. At retrieval, participants performed relative recency judgments for item pairs drawn either within the same event or across the boundary. In the context-present condition, retrieval trials included the original background context; in the context-absent condition, items were presented against neutral backgrounds. A 2 (Context: present vs. absent) × 2 (Boundary: within vs. across) × 3 (Unitization: high vs. medium vs. low) mixed ANOVA on temporal order accuracy revealed significant main effects of Boundary, F(1, 45) = 16.55, p <.001, ηp2 =.07, with within-boundary pairs (M =.70) remembered more accurately than across-boundary pairs (M =.62, p <.001, Cohen’s d =.55); and of Unitization, F(2, 90) = 7.18, p<.001, ηp2=.04, with medium unitization (M =.70) showing significantly higher temporal order accuracy than both low unitization (M =.65, p <.05, Cohen's d =.34) and high unitization (M =.63, p<.01, Cohen's d=.45). Importantly, the joint analysis revealed that in the high unitization condition, boundary effects were eliminated. The accuracy of responses across boundary (M =.66) did not differ from that of responses within boundary (M =.61, p >.05). Although context presentation also improved temporal order memory overall, the three-way interaction was not significant. This suggests that context presentation and encoding unitization operated as relatively independent regulatory mechanisms.
In Experiment 2, 75 undergraduates were randomly assigned to one of three encoding-strategy groups to induce top-down unitization levels: interactive imagery (high unitization), conceptual definition (medium), and item comparison (low). During encoding, participants followed strategy-specific instructions to process each item pair; each sequence again contained an event boundary marked by a change in background context. After a distractor task, temporal order and source memory were assessed for within- and across-boundary pairs without contextual backgrounds to purely examine the effects of encoding strategies. A 2 (Boundary: within vs. across) × 3 (Unitization: high vs. medium vs. low) mixed ANOVA on temporal order accuracy showed a significant Boundary effect, F(1, 72) = 7.28, p <.01, ηp2 =.04, with within-boundary performance (M =.68) exceeding across-boundary (M =.61, p<.01, Cohen’s d =.38), and a robust Unitization effect, F(2, 72) = 15.73, p <.001, ηp2 =.18, reflecting higher accuracy in the interactive imagery group (M =.72) than in the conceptual definition (M =.67) and item comparison groups (M =.55). Critically, a significant Boundary × Unitization interaction, F(2, 72) = 3.29, p <.05, ηp2=.04, indicated that only the interactive imagery group showed no significant boundary effect, demonstrating that effective top-down unitization can neutralize boundary-induced fragmentation.
Across two experiments, we demonstrated that both perceptually driven (bottom-up) and strategically guided (top-down) unitization at encoding significantly reduce the disruptive impact of event boundaries on temporal order memory. Our findings reveal a crucial mechanistic shift from "retrieval support" to "encoding optimization": rather than relying solely on external scaffolding during retrieval, high-quality integrated representations formed during encoding can proactively resist boundary-induced segmentation. Consistent enhancements in source memory accuracy and temporal order improvements provide direct evidence that unitization encoding strengthens the quality of item-context binding rather than merely reinforcing inter-item associations. These results demonstrate the malleability of event boundary effects and identify the modifiable conditions that can optimize temporal memory encoding and retrieval of complex experiences. Future research should explore the potential compensatory interactions between encoding unitization and retrieval context support, and employ neuroimaging to elucidate the neural mechanisms underlying unitization-mediated boundary attenuation. Clinically and educationally, training individuals to form integrated representations and leverage contextual details may enhance memory for temporal sequences in everyday life.
Student mental health has emerged as a critical concern in China's basic education system. The Ministry of Education's comprehensive action plan (2023-2025) emphasizes “Five-Dimensional Education” to promote student psychological well-being. The Gaokao (National College Entrance Examination) essay component serves multiple functions. It implements moral education, facilitates talent selection, and guides instructional practices. Essay prompts exert substantial washback effects on Chinese language teaching. They also strongly influence students’ cognitive development and psychological growth. However, existing research has predominantly focused on moral education, language instruction, and social value orientation. There is limited systematic analysis from a mental health perspective. This study addresses this gap by examining the integration and evolution of mental health elements in Gaokao essay prompts.
The investigation employed a mixed-methods approach comprising two complementary studies. Study 1 utilized semi-structured interviews with 15 Chinese high school teachers (Study 1A) and 15 students (Study 1B). The study explored mental health content in essay prompts. Thematic analysis revealed 6 core mental health dimensions: learning, self-awareness, interpersonal relationships, emotional regulation, social adaptation, and career planning. Study 2 built upon these findings. It developed a coding framework based on the Guidelines for Mental Health Education in Primary and Secondary Schools (2012 revision). Eleven coders with backgrounds in psychology and Chinese language independently analyzed 98 Gaokao essay prompts from 2015-2024. They used back-to-back coding. Statistical analysis was conducted using R language. The analysis encompassed frequency analysis, trend examination, and Representational Similarity Analysis (RSA).
The results demonstrated that Gaokao essay prompts consistently integrate mental health elements across all 6 identified dimensions. Self-awareness and social adaptation emerged as the most frequently addressed themes throughout the decade. The integration patterns responded strongly to educational policies and social issues. There was no significant temporal evolution. Regional variations in thematic emphasis were observed. New curriculum reform prompts maintained continuity with traditional formats in mental health element integration. RSA revealed distinct representational patterns across different regions. New examination formats preserved the core mental health themes from conventional approaches.
These findings indicate that Gaokao essay prompts function as implicit carriers of mental health education. They systematically embed psychological development elements within academic assessment. The prompts demonstrate broad coverage, though the distribution of mental health dimensions was uneven. They show strong temporal responsiveness to societal needs, high regional adaptability, and simultaneous content continuity with format innovation. This study represents the first systematic investigation of mental health element integration in Gaokao essay prompts. It employs both quantitative and qualitative methodologies. The research reveals hidden psychological education functions within China's most influential academic assessment. It provides empirical support for the “Five-Dimensional Education” approach to student mental health promotion.
The study offers practical guidance for explicit construction of mental health elements in essay prompt design and classroom instruction. Ultimately, it contributes to the comprehensive psychological development of students within China's evolving educational landscape.
It is well known that adolescents are prone to risk-taking behavior. Given that previous research on peer influence has focused primarily on peer presence, the importance of peer group competition as another key aspect of peer environment that influences risk-taking behavior has been overlooked. According to the biopsychosocial model (BPS), individual risk-taking behavior results from physiological, psychological and social-environmental factors. As a common social phenomenon, previous studies have shown that competition can lead to individual risk-taking behavior. However, in real life, individuals usually belong to a certain group or social situation, and often participate in comparison and competition in the form of teams. Due to the strong sense of team awareness and collective honor of adolescents, their inter-group competition may also lead to individual risk-taking behavior. Furthermore, individual behavior is also highly dependent on the cognitive assessment of the current situation. Subjective perceptions and evaluations of the potential risks of the external environment can predict the occurrence of individual risk-taking behaviors. When faced with competitive situations and different outcomes, both the belief on whether oneself or the collective can complete a certain task and the cognitive evaluation of the task situation may further affect the individual’s subsequent behavior performance. First, prior success or failure experiences may cause changes in efficacy, including individual self-efficacy and collective efficacy, which in turn affects risk-taking behavior. Second, when faced with competitive pressures, individuals may have two different cognitive responses, challenge and threat, with individuals with challenging perceptions tending to have cognitive states oriented toward approach motivation, and therefore more inclined to take risks; individuals with threatening cognitions, on the other hand, have avoidance motives and thus a low propensity to take risks. Finally, collective efficacy may provide individuals with a form of peer or social support that alters their cognitive appraisal of the task, which in turn influences subsequent performance. When an individual perceives that their team is capable of responding to a crisis (high collective efficacy), their threat perception decreases, resulting in low risk-taking behavior.
Therefore, to explore the effect of inter-group competition on adolescents’ risk-taking behavior and the mediating roles of efficacy and challenge-threat cognition in this process, a total of 274 middle school students (age: M ± SD = 12.84 ±.66 years, 141 girls) were invited to participate. A between-subject design was used, and all participants completed the General Self-Efficacy Scale and the Collective Efficacy Scale (i.e., pre-tests). Then, they entered the classroom, with the experimental group (pink and blue teams wearing the appropriate color bracelets) entered the classroom together one group at a time, and the control group (green team) entered the classroom one group at a time. The experimenter informed the participants that they were to complete the Transfer Table Tennis Match in teams of five. After that, they completed the Challenge-Threat States Appraisal Questionnaire, the General Self-Efficacy Scale, and the Collective Efficacy Scale (post-test) sequentially, and finally completed the Hot version of the Columbia Card Task in the computer classroom.
The results showed that: (1) Inter-group competition did not directly affect adolescents’ risk-taking behaviors. Rather, it affected risk-taking behaviors through the "bridge" of collective and self-efficacy. Adolescents who lost the competition had lower collective and self-efficacy, which led to a greater tendency to take risks, and vice versa for those who won the competition. (2) Contrary to the results of the mediating path of efficacy, collective efficacy and challenge-threat cognition showed a "suppressing effect" on the relation between inter-group competition and adolescent risk-taking behavior. Adolescents with high collective efficacy triggered by winning inter-group competition tended to perceive the competition as a challenge, leading to a high propensity to take risks. The opposite was true in the case of competition failure.
In conclusion, inter-group competition was associated with adolescents’ risk-taking behavior through collective efficacy, self-efficacy, and challenge-threat cognition. Challenge-threat cognition was a key mechanism in inter-group competition and adolescent risk-taking behavior. This study enriches the research field of risk-taking behavior and provides empirical evidence to promote healthy adolescent development.
Traditionally, creativity has been regarded as a positive force that drives social progress and individual development. However, when driven by malicious motivations, creativity may also manifest negative aspects, transforming into malevolent creativity behavior. This behavior refers to acts where individuals consciously employ creative thinking to harm others or satisfy improper self-interests. Such conduct not only violates social norms, but also exerts profound negative impacts on personal development and social order. The high school period represents a critical phase in adolescents' socialization process and the formation of creative thinking. During this stage, while demonstrating significantly enhanced cognitive activity, adolescents often exhibit noticeable rebellious tendencies. Coupled with their susceptibility to multiple influencing factors including negative external environments, cognitive development levels, and individual characteristics, this developmental period becomes particularly crucial for the formation and development of malevolent creativity behavior. Therefore, investigating the developmental characteristics and influencing factors of malevolent creativity behavior among high school students carries significant research importance.
Existing research on malevolent creativity behavior has predominantly employed variable-centered cross-sectional designs, which struggle to reveal the intrinsic heterogeneous categories and dynamic developmental processes. In contrast, person-centered latent transition analysis can effectively identify the potential categories of malevolent creativity behavior and their developmental features. Previous studies have indicated that perceived discrimination among adolescents may exacerbate their aggressive tendencies and other behavioral issues; whereas individuals skilled in cognitive reappraisal techniques and with stronger psychological resilience are generally more capable of positively reframing adverse situations, thereby reducing the generation of hostile cognition. Although high school students' perceived discrimination, cognitive reappraisal, and psychological resilience may be associated with malevolent creativity behavior, their relations to the developmental changes of malevolent creativity behavior remain unclear.
The present study adopted a longitudinal design, administering two waves of questionnaires over a six-month period to 1,757 high school students (including both 10th and 11th graders). In each survey wave, participants completed the Malevolent Creativity Behavior Scale, the Perceived Discrimination Scale, the Cognitive Reappraisal Scale, the Psychological Resilience Scale, and demographic information questionnaire. Data analysis was performed using SPSS 26.0 and Mplus 8.3 software, with specific methods including descriptive statistics, correlation analysis, latent profile analysis, latent transition analysis, and multinomial logistic regression.
The results revealed that: (1) Malevolent creativity behavior among high school students demonstrated heterogeneity, with three distinct categories—low, medium, and high—identified at both time points. (2) The low malevolent creativity group exhibited the highest cross-temporal stability, followed by the moderate and high groups; (3) Over time, individuals in the high group showed a greater propensity to transition toward the moderate or low groups; and (4) Gender, grade level, and perceived discrimination were significant predictors of the latent classes and their transitions. Specifically, boys were more likely than girls to transition into the moderate or high groups; students in the first year of high school were more likely than those in the second year to transition from the moderate to the low group; higher perceived discrimination increased the likelihood of transitioning from the moderate to the high group, while psychological resilience only predicted the latent subtypes at T1; and cognitive reappraisal did not significantly predict any transitions in malevolent creativity behaviors.
In summary, this study identified the potential categories of malevolent creativity behavior among high school students and tracked their dynamic changing characteristics. The findings highlight the significant roles of gender, grade level, and perceived discrimination in determining latent class membership and category transitions, while also underscoring the positive role of psychological resilience in shaping these latent profiles. These findings have important implications for preventing and intervening in malevolent creativity behavior among high school students. They suggest that educators should emphasize cultivating students' psychological resilience and actively reduce their perceived discrimination, thereby effectively curbing the formation and development of malevolent creativity behavior.
In today’s flatter organizational structures, the role of coworkers has become increasingly prominent. Compared with feedback from supervisors, feedback from coworkers has unique advantages. First, in terms of frequency, employees communicate more frequently with each other, which enables them to obtain more feedback from their coworkers. Second, in terms of content, coworkers can provide employees with more feedback information related to social norms based on their own work experience. Meanwhile, due to the similarity in work content and environment, coworkers can often gain a deeper understanding of others’ tasks and roles, enabling them to provide more effective feedback. However, the current feedback research predominantly focuses on the supervisor perspective, ignoring the unique influence of coworkers on employees. Coworker developmental feedback refers to useful or valuable information provided by coworkers that enables employees to improve their work. Therefore, this study examines the impact of coworker developmental feedback on employees’ focus on limitations. This study draws on two functional mechanisms from the feedback intervention theory to further explain the influence process of coworker developmental feedback, namely the task processes and the meta-processes. Additionally, when providing feedback, the characteristics of the feedback provider (coworker) will further influence the employee’s interpretation of the feedback. Therefore, this study considers the warm characteristics of coworkers to provide a more comprehensive understanding of the influence process of coworker developmental feedback. In summary, this study aims to explore how coworker developmental feedback affects employees’ focus on limitations through specific feedback intervention processes.
We tested the research hypotheses using a scenario-based experiment and a field questionnaire survey. The scenario-based experiment (Study 1) preliminarily explored the influence of coworker developmental feedback on employees’ task processes and meta-processes. Study 1 adopted a between-subjects scenario experimental design (high coworker developmental feedback vs. low coworker developmental feedback) and randomly assigned participants to two scenarios. We collected data from 173 full-time employees via an online questionnaire platform. The results of study 1 (N = 173) indicated that coworker developmental feedback enhanced employees’ task processes but did not significantly affect their meta-processes.
The field questionnaire study (Study 2) aimed not only to enhance the ecological validity of Study 1’s findings but also to further explore the outcomes and boundary conditions of the impact of coworker developmental feedback. Study 2 conducted a two-wave questionnaire survey, collecting data from 304 full-time employees in China. At Time 1, employees assessed coworker developmental feedback, perceived coworker warmth, task processes, meta-processes, and provided their demographic information. At Time 2 (two weeks after Time 1), employees assessed their focus on limitations. Study 2 (N= 304) replicated the findings in Study 1, with the exception of the non-significant influence of coworker developmental feedback on employees’ meta-processes. Furthermore, the results of study 2 also showed that: (1) Coworker developmental feedback enhanced employees’ task processes, thereby inhibiting their focus on limitations; (2) Coworker developmental feedback also led to an increase in focus on limitations by improving their meta-processes; (3) Perceived coworker warmth moderated the impact of coworker developmental feedback on employees’ focus on limitations, such that higher level of perceived coworker warmth strengthened the inhibitory effect of coworker developmental feedback on employees’ focus on limitations.
This study also makes several theoretical contributions. First, from the perspective of coworkers, this study discusses the influence of coworker developmental feedback on employees’ reactions and behaviors, and expands the research perspective and positive role of developmental feedback. Second, drawing on the feedback intervention theory, this study elucidates how coworker developmental feedback affects employees’ focus on limitations through task processes and meta-processes, thereby uncovering the specific mechanisms of feedback influence. Third, by introducing the perceived coworker warmth into the realm of developmental feedback research, this study offers a more comprehensive and in-depth understanding of how coworker characteristics affect employees’ responses to feedback.
Morality is a cornerstone of human conduct, and moral identity is a critical component thereof. Moral identity refers to the extent to which being "a moral person" is central to an individual’s self-concept. Previous research has primarily focused on the positive effects of moral identity. However, in real-life contexts, there are numerous instances of individuals with high moral identity engaging in moral transgressions, often having double moral standards characterized by being "lenient toward oneself while strict toward others." To address this, we conducted three sub-studies on the Credamo platform to examine whether moral identity amplifies double moral standards and to explore the boundary conditions of this relationship.
Study 1a (N = 885) employed a questionnaire-based approach to preliminarily examine the relationship between moral identity and double moral standards. Participants first completed the Moral Identity Scale and then rated the acceptability of three moral transgressions, which were committed either by themselves or by others. The results showed that, regardless of the level of moral identity, participants accepted their own transgressions significantly more than those of others, indicating that they had double moral standards. However, the difference in acceptability between self and others’ transgressions was more significant among participants with high moral identity, indicating that they had stronger double moral standards.
To further examine the causal relationship between moral identity and double moral standards, Study 1b (N = 660) employed a 2(moral identity: priming vs. control) × 2(judgment target: self vs. others) between-subjects design. The dependent variable was the acceptability of moral transgressions. First, participants completed a moral identity manipulation task. They were told to copy nine moral trait words (priming group) or neutral words (control group) and use some of these words to write a brief personal narrative. Subsequently, participants evaluated the same moral transgressions used in Study 1a. The results showed that, regardless of whether participants were in the control or priming group, participants accepted their own transgressions significantly more than those of others, indicating that they had double moral standards. However, the difference in acceptability between self and others’ transgressions was greater in the priming group, indicating that participants in the priming group had stronger double moral standards.
To explore the boundary conditions that reduce the double moral standards of individuals with high moral identity, Study 2 (N = 344) employed a 3(self-motivation: self-improvement vs. control vs. self-enhancement) × 2(judgment target: self vs. others) between-subjects design. First, participants completed the Moral Identity Scale. According to their mean scores on the scale, participants who had high moral identity were selected. Next, they underwent a self-motivation manipulation, which involved recalling and analyzing their own past immoral behaviors under different conditions. Finally, participants engaged in a modified dictator game, where they decided how much money (0 to 15 yuan) to take for themselves or others, with the remaining amount to be donated. The amount of money taken was the dependent variable. The results showed that, under the self-improvement condition, participants with high moral identity showed no significant difference in the amount of money taken for themselves versus others, demonstrating no double moral standards. In contrast, under both the control and self-enhancement conditions, those with high moral identity took significantly more money for themselves than for others, indicating the presence of double moral standards. These findings suggest that self-improvement motivation can effectively suppress the double moral standards of individuals with high moral identity.
In summary, this study found that, while moral identity can elevate moral standards, it may also amplify double moral standards. However, self-improvement motivation can mitigate the enhancing effect of moral identity on double moral standards. These findings offer valuable insights for designing and implementing effective moral education programs.
Chinese consumers have demonstrated significant interest in China-chic consumption, a recent phenomenon that blends traditional Chinese culture with modern fashion trends. While this trend reflects cultural confidence and drives new consumption growth, the underlying psychological mechanisms remain underexplored. To promote the development of China-chic, it is necessary to explore which types of consumers are engaging in this consumption and the underlying psychological mechanism driving them. Although past research has suggested that cultural identity can impact consumer choices, it remains unclear whether consumer characteristics, which are known to be crucial factors influencing consumption choices, also affect consumers’ preferences for China-chic consumption.
China-chic represents a fusion of Chinese traditional culture and fashion trends, offering both utilitarian and symbolic value. Previous studies have shown that consumers are motivated to engage in symbolic consumption based on their self-construal and expression needs, and derive satisfaction from brands that offer different symbolic values. Self-construal, which refers to how individuals perceive themselves in relation to others, is a key consumer characteristic that influences their consumption behaviors. Based on the meaning transfer model, we propose that individuals with interdependent self-construal prefer China-chic consumption more than those with independent self-construal because the symbolic significance of China-chic satisfies their needs. Therefore, this study aims to uncover the psychological mechanisms that drive China-chic consumption by examining how self-construal influences consumers' preferences in four studies (N = 736), with a focus on the mediating role of psychological ownership and the moderating effect of identity threat.
Study 1 employed experimental manipulation of self-construal to investigate its impact on consumers' preference for China-chic consumption. Studies 1a (N = 140) and 1b (N = 218) utilized a single-factor (self-construal) between-subjects design. Self-construal was activated through story priming, with narratives featuring a tennis player’s competition and a warrior selecting soldiers. Participants were then asked to complete a product selection task. Study 2 (N = 140) utilized a single-factor (self-construal) between-subjects design to further examine the mediating role of psychological ownership. Participants were primed with self-construal and then completed the purchase intention scale and psychological ownership scale. Study 3 (N = 238) examined the moderating role of identity threat, utilizing a 2 (self-construal: independent vs. interdependent) × 2 (identity threat: high vs. low) between-subjects design. After the self-construal priming task, participants were asked to select the best picture three times and were then informed of a weak or strong association with coolness based on the level of identity threat.
The primary findings of this study include: (1) Consumers with interdependent self-construal show a greater preference for China-chic consumption than those with independent self-construal. (2) Psychological ownership mediates the relation between self-construal and China-chic consumption intention. Specifically, consumers with interdependent self-construal are more likely to generate psychological ownership than those with independent self-construal, which in turn drives their China-chic consumption intention. (3) Identity threat moderates the relation between self-construal and China-chic consumption intention. Specifically, when identity threat is high, the consumption tendency of interdependent self-construal consumers is significantly higher than that of independent self-construal consumers. When identity threat is low, there is no significant difference in the consumption tendency between interdependent and independent self-construal consumers.
This study contributes to the literature by shedding light on the expression preferences of individuals with interdependent self-construal and expanding the research scope of self-construal in the context of China-chic consumption. The findings also uncover the mediating role of psychological ownership and the moderating effect of identity threat, providing a deeper understanding of the psychological mechanisms underlying China-chic consumption behavior. From a practical perspective, the research offers valuable marketing insights to companies in the China-chic industry. It suggests that brands can tailor their strategies to target different self-construal types and leverage psychological ownership to enhance consumer preferences. Future research could explore additional boundary conditions and extend the findings to other cultural contexts or product categories.
Although technological advances have substantially improved productivity, individuals report feeling busier than ever. Subjective busyness, the perception that most of one’s time is occupied by work-related tasks, has become a prevalent psychological state in contemporary society, profoundly shaping work habits and lifestyle choices. Many brands recognize consumers' busy mindset and incorporate “busyness” cues into advertisements to build emotional connections, foster resonance, and drive sales of new products. Brand extension, a strategy in which a company leverages an established brand to launch a new product, depends heavily on consumer perception. A key factor in this evaluation is the fit (or similarity), between the parent brand and the extension across functional, image-based, and value-related dimensions. A high fit (a close brand extension) facilitates the transfer of positive brand associations to the new products, increasing its likelihood of success. In contrast, a low fit (distant brand) presents significant differences between the brand and the product, making it difficult for consumers to transfer existing brand associations to the new product. This often triggers cognitive and psychological resistance, leading to lower evaluations. Notably, brands frequently employ busyness-themed appeals when promoting distant extensions. For example, DuPont, a leading industrial technology brand, expanded its high-performance engineering material Tyvek® into lightweight fashion backpacks and apparel, emphasizing “lightweight, efficient, and easy to travel with” to attract time-pressed urban commuters. Similarly, McDonald's has branched into fashion accessories, promoting its brand through ads that highlight “fast-paced” and “on-the-go” lifestyles, symbolically representing the busy urbanite. However, whether and how consumers' own sense of subjective busyness influences their evaluation of brand extension, particularly for distant extensions, remains underexplored. The related hypotheses were examined across three preliminary studies and five main experiments. Study 1a employed a 2 (subjective busyness: high vs. low) × 2 (extension distance: distant vs. close) between-subjects design to examine the effect of high (vs. low) subjective busyness on the evaluation of both distant and close brand extensions (i.e., Hypothesis 1). Study 1b replicated this effect using different busyness manipulations and product categories. Study 2a adopted a 2 (subjective busyness: high vs. low) × 2 (extension distance: distant vs. close) between-subjects experimental design using real brands, to verify the mediation of cognitive flexibility in the relationship between subjective busyness and brand extension evaluation (Hypothesis 2). To enhance the ecological validity of the findings, Study 2b employed a single-factor (subjective busyness: high vs. low) between-subjects design using a fictitious brand to further examine the mediating role of cognitive flexibility (Hypothesis 2) and to rule out other plausible alternative explanations. Study 3 employed a 2 (subjective busyness: high vs. low) × 2 (busyness attribution: positive vs. negative) between-subjects design to test the moderating effect of busyness attribution (i.e., Hypotheses 3a and 3b).
The findings indicate that (1) Subjective busyness influences individuals’ evaluation of brand extensions. Compared to individuals with low subjective busyness, those with high subjective busyness exhibit more favorable evaluations of distant brand extensions. In the context of close brand extensions, however, no significant difference in evaluation is observed between high- and low-subjective-busyness individuals. (2) Cognitive flexibility mediates this relationship. Specifically, a high level of subjective busyness enhances an individual’s cognitive flexibility, which in turn improves the evaluation of distant brand extensions. (3) Busyness attribution moderates the effect. When individuals attribute their busyness positively, subjective busyness boosts cognitive flexibility, thereby leading to higher evaluations of distant brand extensions. In contrast, when busyness is attributed negatively, the aforementioned effect weakens or even disappears. This research extends the brand-extension literature by introducing subjective busyness as a novel antecedent, clarifying its underlying mechanism, and identifying an important boundary condition. It also offers practical guidance for marketers in designing communication strategies for distant brand extensions.
Digital technologies represented by Enterprise Social Networking Platform (ESNP) have become indispensable in modern enterprise operations and management. ESNP have profoundly reshaped employees’ work environments, enabling them to respond more flexibly to increasingly dynamic external conditions. However, many organizations are frustrated that, despite significant investments in deploying ESNP to enhance employee agility, the outcomes are often underwhelming, and sometimes even counterproductive. This paradox has made it a pressing challenge for managers in the digital economy to determine how to fully leverage ESNP to strengthen employee agility. Although scholars have examined the relation between ESNP usage and employee agility, their findings remain inconsistent. To address these contradictions and guide more effective ESNP utilization, it is essential to theoretically and empirically unpack the mechanisms through which ESNP usage influences employee agility.
Drawing on the job demands-resources (JD-R) theory, this study develops a dual-path model to explain the “double-edged sword” effect of ESNP usage on employee agility through two mediating mechanisms: social support and work overload. It also examines the moderating role of employees’ information technology (IT) competence. To test this theoretical framework, a three-wave questionnaire survey was conducted within a large hotel group, yielding 305 valid responses analyzed using structural equation modeling. The empirical findings reveal that: (1) ESNP usage promotes employee agility by facilitating social support, yet simultaneously hinders agility by inducing work overload; (2) Employees’ IT competence amplifies the positive impact of ESNP usage on social support and strengthens the mediating role of social support in fostering agility; and (3) Employees’ IT competence mitigates the adverse impact of ESNP usage on work overload and weakens its mediating effect on employee agility.
This study makes several important contributions. First, it uncovers the dual nature of ESNP usage in shaping employee agility and elucidates its underlying mechanisms. By reconciling conflicting findings in prior literature, it offers a more integrated theoretical framework and robust empirical evidence for understanding how ESNPs affect employee agility. Second, by revealing the contrasting influences of social support and work overload on agility, this research expands and enhances the exising literature on these two critical factors. Third, by incorporating employee IT competence as a boundary condition, the study enriches the understanding of how individual technological capabilities shape the outcomes of digital platform use. In doing so, it expands the boundary conditions of ESNP research and contributes to a more nuanced understanding of employee IT competence within digital work environments.
Beyond its theoretical contributions, this study offers several practical insights. First, organizations should actively encourage employees to use ESNP to build internal social networks and obtain social support, which can help them adapt quickly to external uncertainties. Second, firms should be mindful of the potential for ESNP-induced work overload and take proactive measures—such as setting reasonable communication norms and workload expectations—to minimize its negative effects on agility. Third, when deploying and managing ESNP, organizations should account for differences in employees’ IT competence and provide appropriate training to enhance their ability to use digital tools effectively.
Despite these contributions, this study has several limitations. First, data were collected through self-reported questionnaires from a single source, which may raise concerns about common method bias; future studies should incorporate multi-source or longitudinal data to enhance robustness. Second, because this study was conducted in China, the generalizability of the findings to other cultural or organizational contexts warrants further examination. Future research could validate the proposed model across different regions and industries. Third, this study focused on employees’ IT competence as a moderator; future work may explore additional boundary conditions, such as organizational culture, leadership style, and digital transformation maturity, to gain a more comprehensive understanding of ESNP’s influence on employee agility.
Besides moisturizing and protecting the eyes, tears serve as potent social signals. They reinforce emotional meaning, shape social interactions, and facilitate prosocial behavior. Empirical research has suggested that emotional tears can increase attentional focus on distressed individuals and enhance the motivation to provide support. Despite these findings, the impact of emotional tears on charitable giving, a significant form of prosocial behavior, remains inconclusive. On the one hand, tears may amplify perceived vulnerability and need, thereby eliciting sympathy and increasing the likelihood of donations. On the other hand, excessive emotional displays may be perceived as manipulative or evoke psychological reactance, commonly criticized as “poverty porn.”
Existing research suffers from three key limitations. First, there is a scarcity of studies that directly examine tears in charitable donation contexts. Second, there is an overreliance on single-mechanism explanations, such as sympathy alone. Third, there is a lack of exploration into moderating factors, such as emotional congruency or contextual framing. To address these gaps, the current investigation employed five experimental studies based on real Tencent Charity donation materials to assess the causal influence of tearful expressions on charitable giving, explore the underlying mechanisms, and test boundary conditions.
Study 1a (N= 122) used a single-factor, two-level between-subjects design to compare tearful vs. neutral facial expressions. The results indicated that participants exposed to tearful images reported significantly higher one-time donation amount (p <.05), monthly donation intention (p =.050), and monthly donation amount (p =.066). Study 1b (N= 122) compared tearful expressions with sad expressions without tears, demonstrating that the presence of tears elicited stronger donation responses in one-time donation intention (p <.05), one-time donation amount (p=.001), monthly donation intention (p <.05), and monthly donation amount (p <.01). This supports that tears add a unique social signal beyond sadness alone. Study 1c (N= 88) extended the findings to an ecologically valid online setting, simulating real donation web pages, and confirmed the superior persuasive impact of tearful imagery over neutral expressions (one-time donation intention (p =.059), one-time donation amount (p =.010), monthly donation intention (p =.041), and monthly donation amount (p<.05).
Study 2 (n = 88) introduced happy expressions as the experimental condition and tested a dual-path mediation model. The results revealed that tearful images significantly increased donation intention compared to happy faces (p <.01), and this effect was mediated serially by heightened sympathy and increased perceived need (β=.40, 95% CI [.14,.67]). This not only provides empirical evidence for the empathy-altruism hypothesis in charitable decision-making, but also offers new evidence for the social signaling hypothesis of tears.
Study 3 (n = 161) employed a 2 (expression: tearful vs. non-tearful) × 2 (situational valence: positive vs. negative) design to examine contextual moderation. The findings showed that the effect of tears on donation was stronger in negatively framed scenarios (e.g., disaster relief or urgent need) than in positively framed ones (e.g., recovery success or gratitude), F(1, 149) = 4.27, p <.05, ηp2 =.03; F(1, 149) = 3.76, p =.054, ηp2 =.03. This suggests that the emotional congruency between the stimulus (tears) and the situational valence enhances persuasion, whereas incongruence may undermine the intended emotional appeal due to cognitive dissonance.
Overall, the theoretical significance of these five studies manifests in two respects. On the one hand, we provide new empirical evidence for the empathy-altruism hypothesis; on the other hand, we reveal how tears function as a social signal to facilitate donation intention, and in doing so offer empirical research support for the signal theory of tears. First, we empirically establish that emotional tears can enhance charitable giving, thus extending the functional understanding of tears as social signals. Second, we integrate the signal theory of tears with the empathy-altruism hypothesis, and identify a novel chain mediation —sympathy and perceived need, which explains how tears operate to motivate donation. Third, we identify contextual boundaries that clarify when and why tearful expressions may be effective or counterproductive.
Practically, this research offers actionable insights for charitable organizations and digital fundraising platforms. The strategic use of tearful images, particularly in negative or urgent contexts, may significantly improve campaign outcomes. However, marketers and nonprofit professionals must remain cautious of emotional overuse and ensure congruency between the emotional tone of images and the framing of charitable appeals. Overall, this work deepens theoretical perspectives on the emotional psychology of giving and provides empirically grounded recommendations for enhancing philanthropic engagement.
In the new era, the dissemination and in-depth interpretation of Wang Yangming’s School of Mind hold significant practical importance for promoting the creative transformation and innovative development of traditional Chinese culture. Although numerous studies have been conducted on Liangzhi, disagreement persists regarding its conceptualization and structure due to a lack of empirical approaches. Therefore, grounded in the inherent connection between contemporary discourse on “Liangzhi” and its philosophical origins, this study aims to systematically investigate the diachronic evolution of the conception of Liangzhi and its modern transformation mechanisms. It specifically focuses on analyzing the evolutionary trajectory and potential stability between the traditional structure of Liangzhi and its contemporary cognitive structure.
In Study 1, paragraphs related to Liangzhi were systematically screened and extracted from The Complete Works of Wang Yangming, relying mainly on an online database with the printed edition serving as a supplementary source. Following a process of group discussion and expert review, the research team performed a content analysis of 219 selected paragraphs and sentences using a specially developed classification framework. This process led to the identification of two core doctrinal dimensions: Benti (original substance) and Gongfu (cultivation through practice). Benti was further reflected in three facets: action (the moral practice of knowing and doing good), affection (the affective endorsement of loving good and hating evil), and conation (the aspiration to become a sage). Based on these findings, Liangzhi was conceptualized as an innate moral standard inherent in human nature, one that must be consciously cultivated through gongfu to evolve from a latent potential to a fully manifested state.
In Study 2, a lexical approach was employed to construct a lexicon of Liangzhi descriptors. Data were collected through word association and open-ended questionnaires administered to 324 and 293 university students, respectively. From the 5,861 responses, 108 high-frequency terms were identified. An additional 108 students then rated the importance of these terms on a Likert scale. Terms with mean ratings below 3.5 points were excluded, resulting in a final set of 92 items that constitute the Liangzhi Lexical Rating Scale for contemporary university students. An exploratory factor analysis (EFA) was first conducted with a sample of 542 participants. Subsequently, an independent additional sample of 922 participants was newly recruited and randomly split into two subsamples of 461 each. A second EFA was performed on one of these subsamples, which yielded a five-factor structure. The other subsample was used for confirmatory factor analysis (CFA), which confirmed the structure and showed good model fit (χ2/df = 3.18, CFI =.91, TLI =.90, RMSEA =.07, SRMR =.05). The final scale with 24 items demonstrated sound internal consistency and reliability, validating the five dimensions of Liangzhi: knowledge, affection, conation, action, and gongfu.
This study provides the first empirical examination of the abstract structure of Liangzhi. A comparison of the two studies reveals that while contemporary university students’ cognitive construct of Liangzhi closely aligns with the traditional framework in the dimensions of affection, conation, action, and gongfu, notable differences emerge in their concrete manifestations. For instance, historically significant norms such as “loyalty to the ruler” have lost relevance in modern contexts due to conflicts with contemporary values like equality and democracy. More fundamentally, the dimension of knowledge of Liangzhi has undergone a significant conceptual shift, moving from an emphasis on innate intuition toward cognitive reflection and rational discernment in moral judgment. These findings reflect an overarching trend of intellectualization and rationalization of Liangzhi in modern society. Furthermore, the multidimensional structure of Liangzhi helps explain the motivational drivers behind moral failure. That is, why people sometimes fail to do good and avoid evil. Additionally, it offers a holistic framework for interpreting Liangzhi as a psychological construct.
Aberrant responses, including random guessing, cheating, and mistakes caused by stress, fatigue, carelessness, or misreading, pose significant challenges in educational and psychological assessments. These responses are particularly problematic in Computerized Adaptive Testing (CAT) compared with traditional paper-and-pencil tests. In CAT, examinees cannot revise their answers, and their ability levels are dynamically estimated based on their responses to individual items. This adaptive nature means that early responses can substantially influence the trajectory of subsequent item selection, making CAT especially vulnerable to the effects of aberrant responses. Once these irregular responses occur, they can distort the ability estimation process, leading to a cascade of inappropriate item selections throughout the test.
When examinees produce aberrant responses, considerable deviations may arise in the estimation of their ability parameters. These distortions disrupt the adaptive testing process and may cause the system to select items that are either too difficult or too easy based on inaccurate estimates. Such misalignment compromises the precision of item selection and diminishes the overall validity of the assessment. As these errors accumulate, they can severely weaken the reliability and accuracy of the CAT process. Therefore, effectively managing aberrant responses is essential to maintaining the integrity and effectiveness of CAT systems.
Robust parameter estimation methods are vital for mitigating the adverse effects of aberrant responses. Such methods seek to minimize the influence of anomalous data, thereby improving the reliability of ability estimation and enhancing the overall accuracy of the test. However, research on robust estimation techniques specifically designed to address aberrant responses in CAT remains relatively limited. Therefore, developing more effective and resilient estimation strategies is crucial for advancing the performance and precision of CAT systems.
To address this limitation, a novel method called Weighted Maximum A Posteriori (WMAP) was proposed. Simulation results show that WMAP substantially enhances the accuracy of ability parameter estimation for examinees exhibiting aberrant responses. Compared with traditional estimation approaches, WMAP effectively mitigates the adverse effects of aberrant data, yielding more accurate ability estimates. Notably, it also improves estimation precision for examinees with normal response patterns, offering a dual advantage. This improvement stems from WMAP’s innovative weighting mechanism, which assigns greater weight to responses from items that better correspond to the examinee’s ability level. This feature is particularly beneficial during the early stages of CAT, when item selection may not yet be fully aligned with the examinee’s true ability.
Further validation with empirical data confirms the robust performance of WMAP in real-world testing contexts. Compared with traditional methods, WMAP is more effective in identifying and correcting biases caused by aberrant responses. By minimizing their adverse impact on ability estimation, WMAP produces more accurate and reliable test results. Its weighting mechanism not only diminishes the influence of aberrant responses but also promotes more appropriate item selection aligned with the examinee’s true ability. Consequently, WMAP enhances both the precision of ability estimation and the adaptability of the CAT system, resulting in more reliable assessment outcomes across diverse testing scenarios.
Beyond its immediate applications, WMAP represents a significant advancement in developing robust methodologies for CAT. Its value goes beyond mitigating aberrant responses. Future research could integrate WMAP into response-time models to gain deeper insights into examinee behavior and further reduce measurement error. Moreover, WMAP could be further adapted to handle more complex patterns of aberrant responding, such as those stemming from emotional fluctuations or cognitive fatigue, thereby extending its applicability and practical value. In parallel, complementary advances in item selection strategies could amplify the benefits of WMAP. Dynamically adjusting selection algorithms can account for potential estimation biases, such strategies can minimize errors and achieve more precise item-examinee matching, even under conditions of aberrant behavior. The integration of WMAP with adaptive item selection approaches would significantly enhance the overall performance of CAT systems, particularly in high-stakes testing environments where precision, reliability, and fairness are paramount.
