Lei Zhang

BackgroundLearning from feedback provided by others is important for navigating complex social environments. Aberrant processing of such social feedback is implicated in the psychopathology of several mental health disorders including social anxiety and depression. Recent psychiatric research increasingly adopts transdiagnostic approaches that examine mechanisms and constructs cutting across traditional diagnostic categories. MethodsHere, we applied reinforcement learning models together with transdiagnostic measures of psychopathology to formally assess social evaluation learning. Human participants (n = 193) completed the Social Evaluation Learning Task, learning whether computer personas ‘liked,’ ‘disliked,’ or were ‘neutral’ toward them. Participants also completed six questionnaires assessing socially relevant psychiatric traits. ResultsComputational modeling revealed that a model with separate learning parameters for positive and negative feedback best accounted for the learning behavior. Exploratory factor analysis identified three transdiagnostic factors. A factor reflecting social avoidance predicted enhanced learning from negative feedback and a reduced positive learning bias. Socially avoidant individuals with heightened negative learning rates additionally showed faster reaction times specifically in positive social contexts. A second factor linking with emotional insensitivity for others predicted a lower choice accuracy in both positively and negatively valanced conditions and was associated with faster reaction times overall. No significant associations emerged for a third factor reflecting depressive and mood-related symptoms. ConclusionsThese results demonstrate how transdiagnostic trait dimensions shape social learning mechanisms through specific behavioral and computational processes.

Mentalization - the ability to infer others’ beliefs and intentions to guide one’s own choices - is a key cognitive function underlying human social interactions. Large language models (LLMs) demonstrate behaviour consistent with humans on theory-of-mind tasks, yet whether these models can guide adaptive behaviour through mentalization is unknown. Here we use two economic games with cognitive computational modeling to uncover the latent strategies underlying mentalization in LLMs. We tested individual LLM agents across four model families, DeepSeek, GPT-4.1, GPT-5 and Gemini 2.0 Flash (N = 2,099), against opponents of varying sophistication and examined whether a prompting strategy designed to elicit strategic reasoning improved performance. We benchmarked results against human participants (N = 251) as a comparative measure. Across both games, LLMs showed clear behavioural and computational signatures of mentalizing that differed markedly by model provider and size. Strategic prompting generally improved performance by inducing more sophisticated reasoning, yet the extent of the benefit differed across the two tasks. Last, GPT-5 agents flexibly adapted their recursive depth of reasoning to increasingly sophisticated opponents, demonstrating superior performance to human participants. Collectively, we demonstrate different capacities for mentalization across LLMs, and highlight cognitive computational modeling as a formal method for assessing comparative intelligence across humans and machines.

Background: Learning from feedback provided by others is important for navigating complex social environments. Aberrant processing of such social feedback is implicated in the psychopathology of several mental health disorders including social anxiety and depression. Recent psychiatric research increasingly adopts transdiagnostic approaches that examine mechanisms and constructs cutting across traditional diagnostic categories. Methods: Here, we applied reinforcement learning models together with transdiagnostic measures of psychopathology to formally assess social evaluation learning. Human participants ( n = 193) completed the Social Evaluation Learning Task, learning whether computer personas ‘liked’, ‘disliked’, or were ‘neutral’ toward them. Participants also completed six questionnaires assessing socially relevant psychiatric traits. Results: Computational modeling revealed that a model with separate learning parameters for positive and negative feedback best accounted for the learning behavior. Exploratory factor analysis identified three transdiagnostic factors. A factor reflecting social avoidance predicted enhanced learning from negative feedback and a reduced positive learning bias. Socially avoidant individuals with heightened negative learning rates additionally showed slower decision times specifically in positive social contexts. A second factor linking with emotional insensitivity for others predicted a lower choice accuracy in both positively and negatively valenced conditions. No significant associations emerged for a third factor reflecting depressive and mood-related symptoms. Conclusions: These results demonstrate how transdiagnostic trait dimensions shape social learning mechanisms through specific behavioral and computational processes.

BACKGROUND Chinese international students studying abroad experience elevated levels of psychological distress due to acculturative, academic, and social challenges, yet they frequently encounter barriers to accessing mental health support. Acceptance and Commitment Therapy (ACT) has demonstrated beneficial effects on psychological flexibility and mental health outcomes across diverse populations. However, evidence regarding the feasibility and acceptability of culturally adapted ACT interventions for Chinese international students remains limited. OBJECTIVE This study aims to evaluate the feasibility and acceptability of a culturally adapted online ACT intervention for Chinese international students in the United Kingdom and to generate preliminary estimates of its effects to inform the design of a future definitive randomized controlled trial (RCT). METHODS This three-arm, parallel-group feasibility RCT aims to recruit 60 Chinese international students from UK universities. Participants will be randomly assigned in a 1:1:1 ratio to 1) a 5-week culturally adapted online ACT group intervention, 2) a time-matched online psychoeducation active control group, or 3) a waitlist control group. Assessments will be conducted at baseline (T0), immediately after the intervention (T1), and at the 1-month follow-up (T2). Primary outcomes include feasibility and acceptability, assessed through recruitment, intervention completion, assessment retention, data completeness, participant satisfaction, and focus group interviews. Exploratory outcomes include psychological flexibility (CompACT), psychological distress (DASS-21), and subjective well-being (PANAS and SWLS). Quantitative feasibility and acceptability outcomes will be analyzed descriptively, while qualitative data will undergo reflexive thematic analysis. Exploratory intervention effects will be examined using mixed-design analyses of variance and effect size estimation. RESULTS Participant recruitment and data collection were completed in November 2025. Statistical analyses had not commenced at the time of manuscript preparation. Results will be reported following completion of the feasibility trial. CONCLUSIONS This feasibility trial will evaluate the feasibility and acceptability of a culturally adapted online ACT intervention for Chinese international students. Its findings will inform the design and conduct of a future definitive RCT. CLINICALTRIAL Chinese Clinical Trial Registry ChiCTR2600124236; https://www.chictr.org.cn/showprojEN.html?proj=307398

Rapid offline reactivation of recently learned experiences support adaptive behaviors in changing environments, but whether positive and negative experiences are differentially reactivated immediately after learning and how this shapes decisions remains unclear. Combining fMRI, multivariate predictive modeling and a probabilistic reward-learning task with learning and transfer stages, we examined how gain- and loss-related experiences are encoded during learning, reactivated during post-learning rest, and influence subsequent choices. We used computational modeling to dissociate gain- and loss-related learned experiences, spanning observable stimulus and outcome information as well as latent expected values and prediction errors. Gain- and loss-related experiences were encoded in dissociable neural expressions and were differentially reactivated during post-learning rest. While gain-related experiences exhibited stronger overall reactivation, loss-related reactivation specifically predicted accurate loss avoidance in the subsequent transfer. Moreover, coordinated engagement of a distributed memory network encompassing the posterior cingulate cortex, middle temporal gyrus and hippocampus increased across post-learning rest periods, consistent with systems-level consolidation supporting adaptive loss avoidance. Together, these findings provide evidence that rapid offline reactivation is valence-sensitive and functionally asymmetric, with negative experiences exerting a disproportionate influence on subsequent avoidance despite weaker reinstatement, a pattern that may have broader relevance for understanding persistent negative thoughts in affective disorders.

Addressing climate change requires a better understanding of environmental behavior. Yet environmental-behavior research has largely focused on single observed outcomes, with limited use of computational models to identify the processes through which these behaviors emerge. Many environmental decisions are also socially interdependent, meaning individuals often must weigh personal costs against benefits for others, future generations, and collective outcomes. Social computational neuroscience provides formal models for examining precisely such value-based and social decisions, making it a useful foundation for environmental research. However, environmental decisions also have distinctive features, as their consequences are often delayed, uncertain, and difficult to observe directly. We therefore propose that repeated, computationally structured tasks can complement conventional environmental experimental designs by estimating latent decision processes rather than only measuring behavioral outcomes. We introduce computational frameworks from social computational neuroscience, focusing on subjective-value, reinforcement-learning, drift-diffusion, and agent-based models, and discuss how to adapt them to study environmental behavior. We illustrate this approach with an eco-product choice task integrating product design, learning from feedback, visual attention, and choice. Overall, we propose social computational neuroscience as a framework for linking individual decision mechanisms to socially situated and collective environmental behavior, contributing to the emerging field of environmental social neuroscience.

Digitalisation has transformed public transport services worldwide, yet disparities in access, skills, and outcomes continue to reinforce a digital divide in how citizens engage with these services. While previous studies have examined the digital divide in internet use and smart technologies, few have explored how these inequalities manifest in everyday interactions with digital public transport platforms, extending to citizen participation in government-provided digital services. To this end, we investigated user experiences of public transport applications in Birmingham, whose digital inclusion is disproportionate to its status as a major urban and economic hub in the UK, by analysing 4,275 user reviews on public transport services collected from the Apple App Store and Google Play Store between 2016 and 2025. Using a mixed natural language processing (NLP) approach combining topic modelling and sentiment analysis, we identified key themes and emotional tones in user feedback. Results revealed three dominant topics: App Performance, Usability, and Service Satisfaction, each conceptually aligned with the access, ability, and motivation dimensions of the digital divide theory. Sentiment analysis showed generally negative evaluations, particularly around technical access barriers and satisfaction issues. By integrating user-generated data with digital divide theory, these results provide novel empirical evidence of how inequalities in digital engagement persist within digital twin planning and smart city transport systems, and highlight the importance of inclusive digital design for equitable and sustainable mobility access.

University placements provide valuable hands-on research experience for students, whilst allowing labs to engage in meaningful outreach. However, building and structuring these placements for college students (ages 16 to 18) may present challenges. Firstly, many hosts lack teaching experience towards this specific group, which differ significantly from university students in their subject knowledge and research skills. Additionally, hosts must balance creating engaging content while ensuring students develop meaningful practical skills and knowledge. Drawing from our collective experience, we provide practical guidelines in structuring placements in the fields of psychology and cognitive neuroscience. We discuss the importance of advance planning, tailoring content to student interests, incorporating practical elements, and promoting interdisciplinary engagement. In doing so, we aim to encourage more labs to host student placements, subsequently widening access for psychology and neuroscience research.