<br> Despite rising concerns about sycophancy—excessive agreement or flattery from artificial intelligence (AI) systems—little is known about its prevalence or consequences. We show that sycophancy is widespread and harmful. Across 11 state-of-the-art models, AI affirmed users’ actions 49% more often than humans, even when queries involved deception, illegality, or other harms. In three preregistered experiments (<br> N</i><br> = 2405), even a single interaction with sycophantic AI reduced participants’ willingness to take responsibility and repair interpersonal conflicts, while increasing their conviction that they were right. Despite distorting judgment, sycophantic models were trusted and preferred. This creates perverse incentives for sycophancy to persist: The very feature that causes harm also drives engagement. Our findings underscore the need for design, evaluation, and accountability mechanisms to protect user well-being.<br> <<<

Abstract<br> <br> Background and Hypothesis<br> Stress sensitivity may represent an important target for resilience-promoting, preventive interventions. Resilience Training (RT) is a 4-session, group-based behavioral intervention, focusing on mindfulness-based skills, that leads to reductions in psychopathology. To investigate the mechanisms of RT, the current study tested whether acquiring the skills taught in RT leads to decreases in psychopathology via reductions in one manifestation of stress sensitivity, emotion reactivity, and associated changes in hippocampal-frontal connectivity.<br> <br> <br> Study Design<br> An open trial of RT was conducted in 103 non-help-seeking young adults with mild-to-moderate psychotic experiences (PEs) and/or symptoms of depression. Transdiagnostic symptoms, emotion reactivity, and mindfulness-related skills were measured, and, in a subset of participants (n = 41), resting-state, functional magnetic resonance imaging (fMRI) data were collected, before and after completion of RT.<br> <br> <br> Study Results<br> Replicating and extending findings of prior studies of RT, significant increases in mindfulness-related RT skills and significant decreases in transdiagnostic symptoms and emotion reactivity, as well as changes in hippocampal-frontal functional connectivity, were observed following RT (all P &amp;lt; .02). Mediation analyses revealed that associations between the acquisition of the RT skills and decreases in symptoms (P &amp;lt; .006) were fully mediated by the decrease in emotion reactivity, which was also correlated with a significant pre-to-post increase in hippocampal-dorsolateral prefrontal cortex connectivity (P = .032).<br> <br> <br> Conclusions<br> RT may lead to improvements in mental health by increasing the capacity to manage day-to-day stress. Future randomized controlled trials with extended follow-up can determine whether such improvements decrease the likelihood of developing disabling mental illnesses in transdiagnostically at-risk individuals.<br> <<<

IntroductionPersistent symptoms after mild traumatic brain injury (mTBI) negatively affect daily functioning and quality of life. Fear avoidance behaviour, a coping style in which people avoid or escape from activities or situations that they expect will exacerbate their symptoms, maybe a particularly potent and modifiable risk factor for chronic disability after mTBI. This study will evaluate the efficacy of graded exposure therapy (GET) for reducing persistent symptoms following mTBI, with two primary aims: (1) To determine whether GET is more effective than usual care; (2) to identify for whom GET is the most effective treatment option, by evaluating whether baseline fear avoidance moderates differences between GET and an active comparator (prescribed aerobic exercise). Our findings will guide evidence-based care after mTBI and enable better matching of mTBI patients to treatments.Methods and analysisWe will conduct a multisite randomised controlled trial with three arms. Participants (n=220) will be recruited from concussion clinics and emergency departments in three Canadian provinces and randomly assigned (1:2:2 ratio) to receive enhanced usual care, GET or prescribed aerobic exercise. The outcome assessment will occur remotely 14–18 weeks following baseline assessment, after completing the 12-week treatment phase. The primary outcome will be symptom severity (Rivermead Post-concussion Symptoms Questionnaire).Ethics and disseminationInformed consent will be obtained from all participants. All study procedures were approved by the local research ethics boards (University of British Columbia Clinical Research Ethics Board, University of Calgary Conjoint Health Research Ethics Board, University Health Network Research Ethics Board—Panel D). Operational approvals were obtained for Vancouver Coastal Health Research Institute and Provincial Health Services Authority. If GET proves effective, we will disseminate the GET treatment manual and present instructional workshops for clinicians.Trial registration numberClinicalTrials.gov #NCT05365776 <<<

Creativity is core to being human. Generative artificial intelligence (AI)—including powerful large language models (LLMs)—holds promise for humans to be more creative by offering new ideas, or less creative by anchoring on generative AI ideas. We study the causal impact of generative AI ideas on the production of short stories in an online experiment where some writers obtained story ideas from an LLM. We find that access to generative AI ideas causes stories to be evaluated as more creative, better written, and more enjoyable, especially among less creative writers. However, generative AI–enabled stories are more similar to each other than stories by humans alone. These results point to an increase in individual creativity at the risk of losing collective novelty. This dynamic resembles a social dilemma: With generative AI, writers are individually better off, but collectively a narrower scope of novel content is produced. Our results have implications for researchers, policy-makers, and practitioners interested in bolstering creativity. <<<