Consciousness and Cognition

Introduction Experiential acceptance refers to the capacity to be open to internal experiences without attempting to change or avoid them. Although acceptance is a core emotion regulation strategy within mindfulness- and acceptance-based interventions (MABIs) and a protective factor for mental health, its conceptualization and implementation remain unclear and ambiguous. The aim of this study was to clarify and develop a comprehensive model of accepting anxiety. Method Twenty-six participants from a non-clinical sample with prior experience in MABIs took part in semi-structured interviews exploring their experience of accepting anxiety. Data collection and analysis followed the principles of Grounded Theory to generate a data-driven model of the acceptance process. Results We identified a five-stage dynamic model involving distinct processes: (Stage 1) observing through the body with attentional focus on interoceptive experience; (Stage 2) identifying and acknowledging anxiety; (Stage 3) validating and normalizing the experience through validation and self-compassion; (Stage 4) not reacting characterized by decentering and nonreactivity; and (Stage 5) staying with the experience via exposure. We also identified facilitating factors that support engagement in the acceptance process. Conclusion These findings refine the understanding of acceptance as a multidimensional emotion regulation process by highlighting an active dynamic involving multiple mechanisms underlying the acceptance of anxiety. This model provides a framework for developing more targeted clinical interventions and for investigating individual and contextual variability in these subprocesses.

Humanity has always admired and created artwork, but the neurocognitive mechanisms behind artistic experience are still elusive. Professional artists and their intimate relationship with their artworks provide a unique opportunity to study the nature of art experience due to their expertise in both art making and art appreciation. During two fMRI tasks, professional artists (N=20) made aesthetic judgments on their own and other artists paintings (aesthetic appreciation task); they also mentally reconstructed the moments when they conceived their artworks or, as a control condition, when they visited now-familiar places for the first time (reconstruction by imagery task). During art appreciation of their own (as compared to other artists) paintings, participants showed stronger recruitment of bilateral posterior parietal cortices, the left lateral occipitotemporal cortex, and the dorso-central sector of the right insula, that is, action-related brain regions also involved in encoding the emotional components of movements. The reconstruction of their own artistic creation (as compared to episodic memory retrieval) involved the left fronto-parietal network associated with motor cognition. Altogether, these results suggest that the mental representations of the actions involved in creating art are integral to the overall artistic experience of painters, supporting an embodied view of the artists experience of art.

Color is a prominent feature of visual experience, yet humans can recognize objects easily and accurately from grayscale images. We examined whether color becomes more useful when spatial information is degraded due to blurring. Participants viewed naturalistic scenes in color or grayscale, and reported whether a named target object was present across a range of blur levels that simulated optical defocus from 0-8 diopters. With unblurred images, performance did not differ between color and grayscale conditions, but as blur increased, recognition accuracy declined. Color provided a modest but reliable advantage at higher levels of blur, suggesting that color becomes increasingly useful when optical quality is degraded. We hypothesize that the evolutionary shift towards trichromacy may have been partially driven by the need to compensate for optical degradation due to aging and/or accumulated light exposure.

Somatic symptoms, such as bodily pain, fatigue, or signs of bodily dissociation, are frequent in the general population, impair mental wellbeing, and form early signs of developing mental disorders, such as depression. Managing somatic symptoms effectively in daily life is a crucial step towards establishing early intervention strategies that prevent the occurrence of mental disorders. Yet, somatic symptoms that occur in daily life have received little scientific attention so far. Here, we ask if mentalizing abilities, specifically the ability to reflect on ones own or others emotion, cognitive, or bodily states, explain somatic symptom burden in daily life. Reflective functioning was assessed in N = 96 healthy individuals via a standardized questionnaire, RFQ-8, in addition to a novel questionnaire focusing on the ability to understand ones own and others bodily reactions, BRFQ-9. Subsequently, over the period of 8 weeks, somatic symptoms were sampled in daily life via a novel Mobile Application that combines standardized questionnaire items of the FFSS, SCL-90, SDQ and SSD-12 with an interactive 3D avatar. 91.7% of participants reported somatic symptoms in the assessment period, and BRFQ scores show a significant negative relationship to overall somatic symptom burden. Such a relationship could not be evidenced for RFQ scores. Body reflective functioning abilities are also a significantly stronger predictor of somatic symptoms and explain more variance than standard reflective functioning abilities. This study introduces a new mobile Application that monitors somatic symptoms in daily life and suggests that body reflective functioning is a novel target for prevention and early intervention techniques with the aim to reduce the negative influence of aberrant bodily feelings on daily life.

Each individual person looks at natural scenes in their own unique way, resulting in a distinct perceptual experience of the world. However, little is known about why such differences in gaze emerge. Here, we test the hypothesis that idiosyncrasies in gaze behavior are predicted by inter-subject variations in internal models--expectations about how scenes typically look. In two experiments, we first characterized participants personal internal models by asking them to draw typical bathroom and kitchen scenes. Individual differences in these drawings were quantified using an objective deep learning pipeline and, in turn, related to individual differences in gaze behavior. In Experiment 1, where participants freely viewed a set of kitchen and bathroom photographs, inter-subject similarities in internal models did not predict inter-subject similarities in gaze. In Experiment 2, we encouraged strategic exploration through gaze-contingent viewing and a memory task. Here, inter-subject similarities in internal models predicted similarities in fixation frequency and the sequence in which different object categories were inspected. These findings suggest that the influence of internal models on visual exploration is stronger under increased sensory uncertainty and when expectation-guided sampling of the environment is encouraged. Together, our results provide new insights into how individual expectations shape gaze behavior and help explain why people differ in how they explore the visual world.

Pretend play is a hallmark behavior in childhood where children create nonliteral meanings. Empirical data supporting the role of social cognition and the decoupling from literality are still scarce during early development. We explored here how the comprehension of pretense affects the visual exploratory behavior of toddlers (n = 44) and adults (n = 65) when they were exposed to short video clips in which an actress performed either real actions (e.g., eating jelly) or pretend actions (e.g., pretending to eat with imaginary food), while varying the complexity of those actions. We analyzed participants exploration of the face in the videos as exploitation of social information. We showed that all observers paid more attention to the face in pretend scenarios than in real ones, measured as longer total looking time in adults and more fixations and revisits to the face in both age groups. We also found more gaze shifts (a measure of information sampling) between the face and the moving hand in the pretend videos in both age groups, mainly at the initial stages of the actions. Additionally, analyses of the scanpaths structure using gaze entropy showed less order in the exploration of pretend videos in both age groups, suggesting that pretense involved greater uncertainty and increased information seeking. The less structured trajectories were observed again mainly in complex pretend scenarios. Taken together, our gaze results indicate that from its developmental origins, the comprehension of pretense relies on social processes linked with information seeking and exploration. Significance StatementDevelopmental theories have long debated whether pretend games are born in conjunction with social capacities in the second year or become integrated later in life. Our study shows that, much like adults, toddlers visually explore pretend scenes gathering more social information and in a less structured manner compared to real-world scenarios, suggesting that the emerging capacity to play with the meaning of things is linked with that of thinking of other minds early in life.

Background Negative and stigmatizing attitudes towards people suffering from mental disorders among healthcare providers often act as a barrier to mental healthcare access. To assess these attitudes in primary care physicians (PCPs), a robust, culturally tailored psychometric tool is crucial. This study aimed to translate and psychometrically validate the MICA-4 to assess negative attitudes among PCPs in Pakistan. Methods We recruited two independent samples of PCPs (n=191, n=329) using non-probability sampling. Three bilingual mental health professionals forward-translated the scale, which was then independently reviewed and back-translated. Cognitive interviews were conducted (n=15 PCPs) to assess comprehension and clarity, for the final version to be used in the study. EFA was conducted on Sample 1 to examine the underlying factor structure of the Urdu MICA-4 items. CFA was then performed on Sample 2 to cross-validate the factor structure identified in Sample 1. Internal consistency and convergent validity were also assessed. Results A three-factor solution was retained, including Views (seven items), reflecting clinicians general evaluative perspectives toward mental illness and professional roles; Stereotypes (five items) representing generalized beliefs and disclosure-related concerns regarding individuals with mental illness, and Stigma (three items) capturing social distancing and perceived threat-related attitudes. The Comparative Fit (CFI = .958) and the Tucker-Lewis Index (TLI = .946) indicated good fit. Three items (9, 13, and 12) were removed due to weak loadings (< .40). Composite reliability ({omega}) indicated adequate internal consistency for the Views ({omega} = .70) and Stereotypes ({omega} = .74) factors, and lower for stigma ({omega} {approx} .53). Convergent validity was modest (.40 to .44). Conclusion The findings support the cautious use of Urdu MICA-4 in Pakistani primary care settings. The variability in the factor structure of the scale across cultures raises a practical implication for its dissemination. When item-level instability repeatedly emerges across contexts, permitting limited, evidence-based refinement may strengthen measurement stability and comparability, as well as its reliability in diverse healthcare settings.

A critical aspect of human cognition is the ability to use our knowledge about the laws of physics to make predictions about physical events. Whether this ability is based on abstract processes or is grounded in our body-environment interactions remains an open debate. We used physical reasoning under altered gravity as a model system to show that humans real-time embodied experience modifies their high-level physical reasoning. Specifically, we tested participants in computerised reasoning games, while disrupting their gravitational signalling using Galvanic Vestibular Stimulation (GVS). Participants failed more and had suboptimal strategies under the GVS condition compared to no-GVS in games requiring reasoning about terrestrial gravity. However, the effects of GVS were reduced when the games included reasoning about altered gravity. Our findings demonstrate how the physical experience of the body shifts high-level cognitive skill as reasoning, suggesting that humans mental representation of the world is grounded in adaptable physical mechanisms.

BackgroundAutonomous sensory meridian response (ASMR) and music-induced frisson are sensory-affective phenomena characterized by tingling, chills, and pronounced emotional responses. Previous research has mainly focused on physiological changes during these experiences, whereas much less is known about whether baseline physiological state is associated with subsequent susceptibility. ObjectiveTo examine whether baseline autonomic flexibility, indexed primarily by heart rate variability (HRV), is associated with later ASMR/frisson responsiveness. Resting EEG measures were included as secondary exploratory markers. MethodsFifteen participants were recruited by convenience sampling; after artifact-based exclusion, 10 participants were included in the analyses. A 5-minute resting baseline EEG and ECG was recorded prior to stimulus presentation. Participants were then exposed to auditory and audiovisual ASMR stimuli, classical music excerpts, and a control stimulus, and reported whether they had experienced ASMR-typical sensations or frisson. Main analyses examined associations between baseline physiological parameters and a combined response-positive outcome. Exploratory analyses included participant-level correlations, comparisons between susceptible and non-susceptible participants, and stimulus-specific effect sizes. ResultsHRV-related measures showed the clearest and most consistent pattern of association with responsiveness. Higher baseline total HRV power was associated with a greater number of response-positive stimuli (r = 0.756, p = 0.011), with similar positive associations for high-frequency HRV (HF; r = 0.672, p = 0.033) and baseline heart rate slope (r = 0.751, p = 0.012). Stimulus-specific analyses likewise showed the most consistent positive baseline effects for total HRV power, with HF and heart rate slope pointing in the same direction. Frontal alpha asymmetry (FAA) was negatively associated with responsiveness ({rho} = -0.862, p = 0.001), but EEG findings overall were less consistent than the HRV-related pattern and are best interpreted as secondary exploratory observations. ConclusionsIn this exploratory pilot sample, baseline HRV, particularly total HRV power, showed the most coherent physiological association with susceptibility to ASMR and music-induced frisson. The findings are consistent with the possibility that these experiences depend not only on stimulus properties, but also on pre-existing physiological state. Given the small sample and exploratory design, the results should be interpreted as hypothesis-generating and require replication in larger confirmatory studies.

When one object approaches another object which, upon touching, moves in the same direction, humans report a vivid impression of one launching the other. Visual adaptation can alter this perception of causality: observers less often report seeing a launch after viewing a stream of launch events. In three experiments, we further characterised how visual adaptation influences the perception of causality by determining the spatial specificity of adaptation and timecourse of recovery from adaptation. In Experiment 1, observers saw ambiguous test events (i.e., the overlap between the two objects varied over trials) at three different horizontal eccentricities. Adaptation was strongest when adaptor and test event were presented at the same eccentricity, and absent when the two were separated by just three degrees of visual angle. Moreover, the perception of causality gradually recovered from adaptation, but remained incomplete. In Experiment 2, both long and short adaptation sequences were highly effective in driving adaptation, and showed no difference in the recovery timecourse, which was complete following more experimental blocks. In Experiment 3, a break without any task-relevant visual input also led to a recovery over the same timespan, but this time, the recovery was instantaneous and incomplete. Altogether, our results provide evidence for highly spatially specific computations, instananeously responding to the onset of adaptation and then gradually recovering from the adaptation over a short time window.

Aim Anxiety disorders in older adults are commonly underdiagnosed and undertreated, especially among minority ethnic groups. This UK multicultural qualitative study aimed to explore and compare beliefs about anxiety among self-reporting White British, South Asian, African and Caribbean older adults, using Leventhal's Common-Sense Model. Methods Individual interviews were conducted with 52 older adults who self-reported anxiety (current or past). Data were managed and analysed using the Framework Method. Professional interpreters facilitated interviews with non-English speakers. Findings The study incorporated the perspectives of 27 older adults with distressing anxiety and 25 with non-distressing anxiety. Participants' beliefs mapped onto the illness-related dimensions in Leventhal's Common-Sense Model. Beliefs about anxiety differed across and within cultural groups, with notable distinctions between participants with distressing and non-distressing anxiety. Those with distressing anxiety neither normalised anxiety nor considered it as an illness trajectory. Overall, participants had a fragmented understanding of anxiety disorders. Specific aspects of older adults' beliefs were influenced by their salient identities rather than their cultural background. Two new dimensions were identified: aggravating factors, believed to trigger or exacerbate anxiety symptoms; protective factors, believed to alleviate or prevent mental health problems. Conclusions Applying Leventhal's Common-Sense Model to anxiety has yielded new insights with significant implications for understanding potential causes of low mental health services use among older adults. Grouping people into broad categories of cultural groups disregards the diversity among individuals within the same group. Cross-cultural research should embrace this diversity and employ nuanced approaches to provide meaningful, person-centred insights into people's perceptions of illness.

Background. Adolescence and young adulthood represent critical developmental stages during which mental disorders often emerge, with the potential to impede perceived quality of life. Spirituality (i.e., the search for the sacred) and self-regulation (i.e., intrinsic processes regulating emotions, thoughts, and behaviors) are recognized as protective factors for mental health. However, their dynamic interplay remains largely unexplored, particularly in real-life and in real-time among youths. This study, developed with the help of young partners, addresses this gap by investigating the longitudinal associations between spirituality, self-regulation, and mental health using an ecological momentary assessment (EMA) approach. Methods and analysis. We plan to recruit 120 adolescents and young adults (aged 16 to 20, expected attrition rate of 20%) from the community to complete a qualitative semi-structured interview assessing their beliefs, spiritual or religious activities, role models, and meaning in life. In addition, participants will take part in a multi-wave intensive longitudinal study. Trait-level assessments will be conducted at two time points, three months apart, to capture between-person differences. Additionally, to assess within-person dynamics, participants will complete EMA surveys four times daily over 10 consecutive days in two waves, also three months apart. Measures will include facets of spirituality (e.g., beliefs, meaning, collective consciousness), self-regulation (e.g., self-control, emotional regulation, impulsivity), as well as mental health indicators (emotional and behavioral symptoms) and quality of life. Qualitative data will be analyzed through a thematic analysis method, whereas quantitative associations will be assessed using Linear Mixed Models (LMM) and network analyses. Ethics and dissemination. Ethical approval has been obtained, and data collection begun in May 2025. Findings will be disseminated through open access peer-reviewed journals, conferences on adolescent mental health, and shared with practitioners, educators, and youth organizations. Results will also be made accessible to the general public. This study aims to inform personalized preventive and therapeutic interventions by elucidating real-time mechanisms linking spirituality, self-regulation, and mental health in youths.

Decisions based on evidence accumulated over time require rules governing when to end the accumulation process and commit to a choice. These rules control inherent trade-offs between decision speed and accuracy, which require careful balance to maximize quantities that depend on both like reward rate. We previously showed that, to maximize reward rate, normative decision rules adapt to changing task conditions (Barendregt et al., 2022). Here we used a novel task to examine whether and how people use adaptive rules for individual decisions under a variety of conditions, including changes in decision outcomes across trials and changes in evidence quality both across and within trials. We found that the participants tended to use rules that adjusted, at least partially, to predictable changes in task conditions to improve reward rate, consistent with a rationally bounded implementation of normative principles. These findings help inform our understanding of the extent and limits of flexible decision formation in the brain.

AIM Older adults experiencing anxiety disorders, particularly those from minority ethnic backgrounds, are less likely to use formal mental health services compared to their younger counterparts. This UK multicultural qualitative study aimed to explore and compare beliefs underpinning coping strategies for anxiety among self-reporting White British, South Asian, African and Caribbean older adults, using Leventhal's Common-Sense Model of Self-Regulation. METHODS Individual semi-structured interviews were undertaken with 52 older adults aged 65 and over who self-reported (current or past) anxiety. Professional interpreters supported interviews with non-English-speaking participants (n=10). Eight public contributors collaborated on different aspects of the study. The Framework Method was used to manage and analyse the data. FINDINGS The study drew on the perspectives of 27 older adults with distressing anxiety and 25 with non-distressing anxiety. Across all cultural groups, participants adopted different strategies to manage anxiety, the most prominent of which were self-help strategies. Help-seeking behaviour was influenced by a complex interplay of factors not recognised by Leventhal's Common-Sense Model. Notably, older adults' salient identities, rather than their cultural backgrounds, influenced their selection of coping strategies. CONCLUSIONS Interventions that empower older adults to use self-help strategies more effectively can serve as acceptable adjuncts to formal therapy. Nevertheless, addressing barriers to formal help-seeking is essential, particularly among those with a perceived need to seek help. No one model can depict the complexity of coping behaviours. While applying Leventhal's Common-Sense Model yielded novel insights, it could not fully capture the motivational factors underlying participation in specific coping behaviours. To provide nuanced and accurate insights, cross-cultural research should acknowledge heterogeneity within groups rather than impose boundaries of purportedly homogeneous entities.

Listeners face many challenges when trying to maintain attention to a target source in everyday settings; for instance, reverberation distorts acoustic cues and interruptions capture attention. However, little is known about how these challenges affect the ability to maintain selective attention. Here, we measured syllable recall accuracy and pupil dilation during a spatial selective attention task that was sometimes disrupted. Participants heard two competing, temporally interleaved syllable streams presented in pseudo-anechoic or reverberant environments. On randomly selected trials, a sudden interruption occurred mid-sequence. Compared to anechoic trials, reverberant performance was worse overall, and the interrupter disrupted performance. In uninterrupted trials, reverberation reduced peak pupil dilation both when it was consistent across all stimuli in a block and when it was randomized trial to trial, suggesting temporal smearing reduced clarity of the scene and the salience of events in the ongoing streams. Pupil dilations in response to interruptions indicated perceptual salience was strong across reverberant and anechoic conditions. Specifically, baseline pupil size before trials did not vary across room conditions, and mixing or blocking of trials (altering stimulus expectations) had no impact on pupillary responses. Together, these findings highlight that stimulus salience drives cognitive load more strongly than does task performance.

Reaction time is a measure of the speed of our response to stimuli in the environment. Even for a well-trained task, a subjects reaction time varies. One source of this variability is internal state fluctuations (such as changes in arousal). There are few studies that systematically quantify the extent to which reaction time varies across different timescales and link this to measures of systemic physiology associated with arousal. In much of the literature, it is assumed but not demonstrated that behavioral and systemic measurements associated with arousal will be consistently linked because both estimate a common underlying arousal process. In this work, we examined this assumption by simultaneously measuring reaction time, heart rate, and pupil diameter in rhesus macaque monkeys performing several visual tasks over hours and across hundreds of sessions. We found a portion of the variability in reaction time could be linked to systemic physiological signatures of arousal on fast timescales from second to second and slower timescales from minute to minute. This link between reaction time and systemic physiology was also present for different biomarkers of arousal (heart rate and pupil). However, the strength of this relationship varied depending on the arousal biomarker. Our findings support the conclusion that there are multiple arousal mechanisms that act simultaneously to influence behavior and multiple timescales at which they operate.

Some images are consistently remembered better than others, suggesting that memorability reflects intrinsic image properties. We tested whether within-category distinctiveness underlies this effect. Across three experiments (N = 477), participants categorized indoor scenes previously rated for subjective typicality and then completed recognition memory tests. Typical scenes were categorized faster and more accurately, but were remembered worse and showed a more liberal response bias than atypical scenes. These opposing effects were robust across categories. To link subjective typicality to visual representations, we quantified image distinctiveness using a convolutional neural network (CNN). Across layers, CNN-derived distinctiveness closely tracked human typicality judgments and predicted both categorization speed and memorability, with strongest effects in higher, semantic layers. Critically, the memory advantage for atypical scenes persisted even when most images were atypical, ruling out rarity within the experimental context. Together, the results show that intrinsic scene memorability reflects an images position within a category-specific representational space.

ABSTRACT Background Liberian and Sierra Leonean children born during and after the 2012 UNHCR cessation clause, and the subsequent closure of the Oru refugee camp in Nigeria, have grown up in conditions of protracted displacement and de facto statelessness. Many of these children have been exposed to multiple forms of trauma, including witnessing violence as well as physical, emotional, and sexual adversities within a complex and resource-constrained environment. Many also experience cultural-linguistic disruptions, including heritage-language attrition and increased reliance on host-country languages, which may be associated with challenges in identity formation and social integration. However, little is known about how trauma exposure interacts with language-related factors to influence PTSD and complex PTSD (CPTSD)-related functional impairment among stateless refugee children. Methods Using a cross-sectional design, 320 children aged 6-17 years (180 Liberian, 140 Sierra Leonean) were assessed. Trauma exposure was measured using the Child and Adolescent Trauma Screen (CATS), and PTSD/CPTSD functional impairment using the International Trauma Questionnaire-Child and Adolescent Version (ITQ-CA). Heritage- and host-language proficiency were assessed using a structured sociolinguistic questionnaire. Multivariate covariance analyses were conducted using SPSS to examine main and interaction effects. Results Multivariate analyses revealed that poorer host-language communication was associated with higher PTSD-related functional impairment (F(3, 311) = 2.85, p = .038, partial eta-squared = .027), whereas CPTSD impairment was largely unaffected. Native-language proficiency also predicted PTSD impairment (F(3, 290) = 3.44, p = .017, partial eta-squared = .034), and children with low heritage-language skills, limited parental/home-language exposure, and no Nigerian-language use showed the highest CPTSD impairment. Emotional connection to the native language provided a modest protective effect. The combined heritage- and host-language exposure was linked to lower trauma-related functional impairment, particularly for children at higher risk of CPTSD. Witnessed trauma emerged as the strongest predictor of functional impairment among refugee children, with CPTSD outcomes showing greater sensitivity (partial eta-squared = .153) than PTSD (partial eta-squared = .076). Conclusions Heritage-language competence and bilingual proficiency were associated with reduced PTSD-related functional impairment, whereas CPTSD was more strongly shaped by cumulative relational trauma. These findings highlight the potential value of interventions that support bilingual development and heritage-language preservation as pathways to resilience among stateless refugee children. Keywords: Language attrition; bilingual competence; trauma exposure; refugee children; CPTSD; PTSD; functional impairment

BackgroundAI-powered cognitive behavioural therapy (AI-CBT) tools hold significant promise for addressing the global mental health treatment gap, yet sustained user engagement remains critically low. While patient attitudes and experiential factors have been qualitatively documented, the psychological mechanisms through which AI literacy translates into long-term engagement remain poorly understood. Existing systematic evidence highlights trust, perceived therapeutic alliance, and stigma as salient themes, but no large-scale quantitative study has modelled these as a mediated pathway. ObjectiveThis study aimed to (1) examine whether trust in AI systems and perceived therapeutic alliance mediate the relationship between AI literacy and sustained AI-CBT engagement, and (2) determine whether mental health stigma moderates these mediated pathways. MethodsA cross-sectional national online survey was conducted in the United Kingdom (N = 1,247). Eligible adults (18+) with a history of anxiety or depression who had used an AI-CBT tool in the preceding 12 months were recruited via stratified random sampling. Structural equation modelling (SEM) with moderated mediation was conducted in R (lavaan 0.6-17). Moderated mediation was evaluated using the PROCESS macro framework adapted for SEM, with 5,000 bootstrap replications for bias-corrected confidence intervals. Model fit was assessed using CFI, TLI, RMSEA, and SRMR indices. ResultsThe final SEM demonstrated excellent fit (CFI = 0.967, TLI = 0.959, RMSEA = 0.043 [90% CI: 0.036-0.051], SRMR = 0.052). AI literacy exerted a significant indirect effect on sustained engagement through trust in AI ({beta} = 0.213, SE = 0.031, p < .001) and perceived therapeutic alliance ({beta} = 0.187, SE = 0.028, p < .001). Mental health stigma significantly moderated the trust[-&gt;]engagement pathway ({Delta}R2 = 0.042, p = .003), with the indirect effect being stronger among individuals with lower stigma scores. The total indirect effect accounted for 58.4% of the total effect of AI literacy on engagement. ConclusionsAI literacy promotes sustained AI-CBT engagement primarily through its effects on trust and perceived therapeutic alliance, pathways that are attenuated by mental health stigma. These findings underscore the need for stigma-reduction interventions and AI literacy programmes as implementation strategies. Findings have direct implications for the design and deployment of AI-CBT tools across UK NHS digital mental health services.

BackgroundMillions of people use language models to discuss mental health concerns, including suicidal ideation, but limited frameworks exist for evaluating whether these systems respond safely. Benchmarking, the practice of administering standardized assessments to language models, offers direct parallels to clinical competency evaluation, yet few clinicians are involved in designing, validating, or interpreting these assessments. AimsTo introduce mental health professionals to benchmarking language models by administering a validated clinical instrument and demonstrating how configuration decisions, measurement limitations, and scoring context affect result interpretation. MethodWe administered the Suicide Intervention Response Inventory (SIRI-2) programmatically to nine commercially available language models from three providers. Each item was presented 60 times per model (three prompt variants x two temperature settings x 10 repetitions), yielding 27,000 model responses compared against point-in-time expert consensus. ResultsTotal scores ranged from 19.5 to 84.0 (expert panel baseline: 32.5). Prompt design alone shifted individual model scores by as much as the difference between trained and untrained human groups. The best performing model approached the instruments measurement floor. All nine models consistently overrated clinically inappropriate responses that sounded supportive. ConclusionsA single benchmark score can support markedly different claims depending on the assumed standard of clinical behavior, the instruments remaining measurement range, and the configuration that produced the result. The skills required to make these distinctions must become core competencies. Benchmark results are increasingly utilized to support claims about mental health safety that may not be accurate, making it necessary to close the gap between clinical measurement and AI. Plain Language SummaryAI chatbots like ChatGPT, Claude, and Gemini are increasingly used by millions of people to discuss mental health problems, including thoughts of suicide. To assess whether these systems handle such conversations safely, researchers give them standardized tests called benchmarks and compare their answers to those of human experts. These scores are already used to argue AI systems are ready for clinical use. This study gave a well-established test of suicide response skills to nine AI models from three major companies under varying conditions. We changed how much instruction the AI received and how much randomness was built into its responses, then measured whether the scores changed. The same AI model could score like a trained crisis counselor under one set of conditions and like an untrained undergraduate under another, depending on choices the person running the test made. Every model also made the same kind of mistake: responses that sounded warm and caring were rated as appropriate, even when experts had judged them to be clinically problematic. The highest-scoring model performed so well that the test could no longer measure whether it was truly skilled or had simply exceeded the tests range. These findings show that a single score can be misleading without knowing how the test was run, whether it can still distinguish strong from weak performance, and whether it matches what the AI is used for. Mental health professionals routinely make these judgments about clinical assessments and are well positioned to bring that expertise to AI evaluation.