Social anxiety disorder affects approximately 12% of Western populations. STePS-Ai compares AI-delivered cognitive behavioral therapy (CBT) and psychodynamic therapy (PDT) for social anxiety in a randomized controlled trial. The AI chatbots are programmed to deliver either CBT or PDT protocols, though their adherence to these therapeutic frameworks remains to be evaluated. This study builds on findings from the initial STePS project showing minimal differences between therapist-guided and self-guided interventions.

Ninety participants (LSAS-SR ≥30) are randomized 1:1:1 to AI-delivered CBT, AI-delivered PDT, or waitlist control. Both active treatments consist of 4-week smartphone-based interventions where an AI assistant provides daily text-based conversations through end-to-end encrypted applications. The Social Phobia Inventory (SPIN) serves as the primary outcome measure, with additional assessments including LSAS-SR, PHQ-9, GAD-7, Session Alliance Inventory, and Client Satisfaction Questionnaire. Analysis follows intention-to-treat principles using linear mixed models, with response defined as ≥30% reduction in SPIN scores.

The study is currently ongoing with the 4-week treatment phase and post-assessment expected to be completed by the conference date, allowing presentation of preliminary findings on the comparative effectiveness of these approaches.

This preliminary investigation contributes incrementally to the expanding literature on AI-delivered psychological interventions. Findings should be interpreted with caution given methodological limitations and the need for replication. Future research must systematically examine therapeutic alliance formation, treatment fidelity, and comparative effectiveness against traditional delivery methods across diverse clinical populations. While potentially informative regarding treatment modality differences, results represent only one step in the empirical evaluation of AI-assisted psychological interventions for social anxiety disorder.

This pilot study aimed to investigate the acceptability and effectiveness of a brief-guided internet-based cognitive behavioural therapy (iCBT) for people with emotional distress during the COVID-19 pandemic. Using a quasi-experimental research design with a 3-arm study, eligible participants were assigned to a group with student counsellors (n = 24), a group with counsellors (n = 23) or a non-active control group (n = 21). Participants received eight online modules and regular support via telephone counselling or video conferencing app (Zoom) from a counsellor or student counsellor during a 5-week intervention period, while the control group did not receive any intervention during the intervention period. An adherence rate of 85.10% was observed. The results of the 3 (group) × time (pre vs. post) repeated-measures analysis of covariance showed that the student counsellor and counsellor groups demonstrated significantly greater reductions in total emotional distress (partial η² = 0.10) and stress (partial η² = 0.14) than the control group. In addition, student counsellors and counsellors produced comparable intervention effects, and telephone counselling and Zoom support produced comparable intervention effects. This pilot study supports the acceptability and effectiveness of brief-guided iCBT for people with emotional distress during the COVID-19 pandemic.

Introduction: Fear of contamination is one of the most reported symptoms among individuals with obsessive-compulsive disorder (OCD). Digital interventions may address treatment gap caused by geographical constraints, long waiting lists, or reluctance to seek help. Yet, involvement of people with lived experience is necessary for effective implementation and sensible use of resources in health care innovations. We developed ZeroOCD, an augmented reality exposure therapy smartphone application for fear of contamination, while involving experts-by-experience in the whole development, from the first draft to user testing of the final version of the application.

Methods: Following the co-creation framework, we conducted focus groups with eight experts-by-experience and four OCD therapists to address their needs and preferences. Upon the development of the app, we conducted user testing with the same sample to address user-friendliness. We consequently conducted a feasibility study involving 20 participants who followed the intervention with therapeutic support on request.

Results: Based on the insights of the participants, we iteratively co-created nine modules of the app, and the content of the augmented reality exposure. User-centered methodology enabled us to identify topics the participants were particularly interested in, such as specific exposure focus, accessible therapist involvement, or flexible gamification-based reward system, address design challenges, and improve user satisfaction.

Conclusions: Employing co-creation framework involving experts-by-experience may identify unexpected challenges in the intervention design and shed light on content necessary to improve user satisfaction. Thus, it is highly beneficial in the development of mental health innovation and may lead to adherence improvement.

Background: Social anxiety is a heightened fear and discomfort in social situations. Cases of elevated distress and impaired functioning can lead to a clinical diagnosis of social anxiety disorder. Altering cognitive biases associated with social anxiety has been suggested as potentially beneficial; however, little is known about the comparative effectiveness of such interventions. The aim of this living systematic review is to examine the efficacy of cognitive bias modification for reducing social anxiety.

Methods: We will search multiple electronic databases for randomised controlled trials evaluating the efficacy of cognitive bias modification for people diagnosed with social anxiety and people exposed to a social stressor. The primary outcome will be change in social anxiety related symptoms; secondary outcomes will be changes in social functioning and quality of life and adverse events. Study selection, data extraction and risk of bias assessment will be done by at least two reviewers using pre-defined tools. We will synthesise data from people with social anxiety diagnosis and those subjected to a simulated social stressor separately using random effects meta-analyses. Heterogeneity will be evaluated by investigating characteristics of included studies and we will conduct a network meta-analysis in order to compare the efficacy of subtypes of cognitive bias modification for social anxiety disorder. We will appraise the strength of the evidence for each outcome by reviewing the overall association, internal and external validity, and reporting biases. Where data allows, we will triangulate the evidence from both sources with a multidisciplinary group of experts. We will also descriptively report factors reported to mediate cognitive bias modification, The review will begin in living mode and the database search will be rerun every three months to identify potential new evidence. We will co-produce this review with members of a global lived experience advisory board.

This protocol was registered on 15.10.2024 (CRD42024601380).

Background: Reports of artificial intelligence (AI) chatbots fueling delusions in vulnerable users have popularized the notion of “AI psychosis”. We argue the risk is not unprecedented. Individuals with psychosis have long incorporated books, films, music, and emerging technologies into their delusional thinking.

Methods: We review historical parallels, summarize why large language models (LLMs) may reinforce psychotic thinking via sycophancy (excessive agreement or flattery to avoid confrontation), and provide two vignettes contrasting unsafe and safe responses.

Results: Contemporary LLMs often avoid confrontation and may collude with delusions, contrary to clinical best practice.

Conclusion: The phenomenon is not new in principle, but interactivity potentially changes the risk profile. Clinically aware LLMs that detect and gently redirect early psychotic ideation, while encouraging professional help seeking, could reduce harm. Design should be guided by therapeutic principles and evidence about current model failures.

The use of AI in digital mental healthcare promises to make treatments more effective, accessible, and scalable than ever before. At the same time, the use of AI opens a myriad of ethical concerns, including the lack of transparency, the risk of bias leading to increasing social inequalities, and the risk of responsibility gaps. This raises a crucial question: Can we rely on these systems to deliver care that is both ethical and effective? In attempts to regulate and ensure the safe usage of AI-powered tools, calls to trustworthy AI systems have become central. However, the use of terms such as “trust” and “trustworthiness” risks increasing anthropomorphization of AI systems, attaching human moral activities, such as trust, to artificial systems. In this article, we propose that terms such as “trustworthiness” be used with caution regarding AI and that when used, they should reflect an AI system's ability to consistently demonstrate measurable adherence to ethical principles, such as respect for human autonomy, nonmaleficence, fairness, and transparency. On this approach, trustworthy and ethical AI has the possibility of becoming a tangible goal rather than wishful thinking.

As e-health offerings rapidly expand, they are transforming and challenging traditional mental health care systems globally, presenting both promising opportunities and significant risks. This article critically examines the potential and pitfalls of integrating digital technologies into mental health care, particularly in the realms of diagnosis, prevention, and treatment. It explores current advancements and evidence-based practices, and provides a vision for how future technologies can evolve responsibly to meet mental health needs. The article concludes with the TEQUILA framework, addressing essential elements and challenges for fostering a beneficial and ethical future. A responsible future for digital mental health requires building Trust by ensuring data privacy, security, and transparency in AI-driven decisions, along with Evidence-based and robust regulatory oversight to maintain Quality. Usability, design, usability tailored to diverse needs, and ethical alignment with users' Interests will all be essential, while Liability and Accreditation standards will safeguard accountability in this evolving landscape.

Objective: This study aimed to evaluate the feasibility and efficacy of a cognitive behavioral therapy (CBT) group for left-behind older parents with depression. Method: Using a quasi-experimental research design with a three-arm study, participants with depression (n = 17) and participants without depression (n = 13) received the same eight-session CBT group covering the skills of acceptance and commitment therapy in addition to treatment-as-usual (TAU), that is, interest classes and leisure activities, while the control group (n = 5) received TAU. Results: 3 (group) × 2 (time) repeated measures of ANOVA showed that the CBT group for left-behind older adults with depression, as compared with the control group, showed a significantly greater improvement in depression (Cohen's d = 1.0), stress (Cohen's d = 1.1) and subjective well-being (Cohen's d = 0.7). Conclusion: The study results support the feasibility and efficacy of the CBT group for left-behind older adults with depression.

Purpose: This study examined the efficacy of a 10-week internet-based cognitive behavioral therapy (iCBT) program “REST Online” for Hong Kong university students with mild to moderate levels of psychological distress. Method: A total of 206 Hong Kong university students were randomized into: (1) web-based and (2) app-based iCBT, and (3) waitlist control (WLC) groups. Results: Compared with the WLC group, the participants in the two iCBT groups showed a significant reduction in psychological distress, depression and anxiety symptoms, and negative thoughts and emotions, and significant increase in positive thoughts and emotions, with medium to large effect sizes. The positive effects were sustained at the 3-month follow-up test. No significant intervention effects were found between the two iCBT groups except for anxiety symptoms. Discussion: The findings were discussed in terms of program design and service delivery, and suggestions for delivering digital mental health service in Hong Kong universities were proposed.