Open research questions in Digital Mental Health Interventions

What the literature leaves open

• AI-supported how In sum, whether an AI tool is used by a primary care provider to support decision making, or directly by patients for health trust systems in health (i.e., beliefs education, its adoption is likely to be influenced by confidence and that healthcare organizations use AI safely, effectively and responsibly), trust in brokers (i.e., confidence in physicians as appropriate and competent intermediaries in the use of AI and expected benefits (i.e., perceptions that AI in healthcare will improve population health outcomes). These factors are also expected to shape, and be shaped by, comfort with AI use in healthcare. Figure 1 presents the conceptual framework guiding this study and illustrates the hypothesized relationship among these factors and comfort with physician versus patient use of AI for diabetes prevention. Patient-facing AI tools can expand reach and engagement. However, despite the growing use of AI in healthcare, little is known about public and patient comfort with these tools (12), particularly in primary prevention. Existing evidence rarely distinguishes between AI used by physicians to support prevention and AI used directly by patients, leaving gaps in understanding how comfort varies by user and context. This manuscript addresses this gap by examining the use of AI in diabetes prevention, comparing comfort levels when AI tools are used by physicians them independently. We also assess whether and how demographic and other predictors play a role in comfort in each context. This study explores these questions in primary prevention settings, offering insights with implications for policy, practice, and implementation science. Understanding these interactions can help providers, health systems, and policymakers determine the conditions under which AI tools for primary prevention are most acceptable, and how individual and contextual factors shape comfort with their use.

  Physician versus patient use of AI for diabetes prevention: public perceptions and comfort levels · 2026 · DOI

• Catharina F. van der Boor1 and Paul E.W. van der Boor2 Cite this article: van der Boor CF and van der Boor PEW (2026). The humanitarian AI paradox: Key opportunities, challenges and research needs for the use of AI in humanitarian mental health response.

  The Humanitarian AI Paradox: Key Opportunities, Challenges and Research Needs for the use of AI in Humanitarian Mental health Response · 2026 · DOI

• Future research should address these concerns and evalu- ate whether automated text messaging-based interventions can effectively improve suicide prevention outcomes while remaining acceptable and engaging to young adults. Their perspectives may not be representative of the broader young adult population with lived experience of suicidal thoughts. Second, our results are representative of young adult MHA users who experienced suicidal thoughts and might not be generalizable to all young adults who expe- rience suicidal thoughts. Finally, the ideas and opinions expressed in this study are extracted from stated, as opposed to demon- strated, preferences about a hypothetical digital suicide pre- vention and past experiences with suicide-related thoughts and behaviors and may not be representative of how young people with suicidal thoughts make use of or want to inter- act with a digital suicide prevention intervention.

  Exploring Barriers to Crisis Support: Considerations for the Design of Automated Digital Safety Planning Interventions · 2026 · DOI

• First, the bibliometric analysis is limited by the databases and search criteria used, which may have led to the exclusion of studies indexed in other databases or published in other languages. Third, bibliometric methods primarily focus on quantitative indicators such as the number of publications, citations, and collaboration networks, thereby providing limited information on the methodological quality or clinical efficacy of the studies.

  The Rise of Artificial Intelligence in Cognitive Behavioral Therapy Research: A Global Bibliometric and Network Analysis · 2026 · DOI

• Taken together, these findings indicate that current systems remain insufficiently adapted to the MENA context, underscoring the need for culturally grounded, dialect-sensitive, and clinically supervised approaches to ensure safe and effective integration.

  From promise to practice: artificial intelligence in mental health care in the MENA region · 2026 · DOI

• Building the foundations for global data banking in digital phenotyping for mental health Bridianne O’Dea, Samra Naz1, Larisa T. McLoughlin1 and Alexis E. Whitton 2 1,2 ✉ © The Author(s) 2026:,;) ( 0 9 8 7 6 5 4 3 2 1 Mental illness is a leading cause of global disability, underscoring the urgent need for scalable, data-driven approaches to early identification and intervention. Passive sensing technologies in mobile and wearable devices enable continuous and unobtrusive measurement of behavioural and physiological signals that may be relevant to mental health. When translated into interpretable digital markers through digital phenotyping, these data hold significant promise for advancing our understanding of the onset, course, and treatment of mental disorders. However, achieving real-world clinical utility requires large-scale, harmonised, and ethically governed databanks that enable replication and generalisability for diverse populations. Informed by recent international initiatives, the academic literature, and our own expertise in digital phenotyping, this Perspective outlines four key priorities for advancing digital phenotyping databanks in depression and anxiety. First, ensuring data quality through standardisation and harmonisation is essential to comparability across studies and to prevent fragmentation. Second, ethical data stewardship demands hybrid consent models that combine the scalability of broad consent with the flexibility of dynamic consent, ensuring meaningful participant control as analytics evolve. Third, robust, privacy-preserving information governance co-created with people with Lived Experience is vital to maintain trust and prevent misuse, with federated learning and open-source pipelines offering promising technical pathways. Finally, the field must promote data reuse by reforming incentive structures, recognising databank-based scholarship, and investing in sustainable infrastructures that reward secondary analyses. Collectively, these priorities offer a pragmatic framework for building equitable, transparent, and scientifically robust digital mental health databanks. Implementing these recommendations will require sustained international collaboration among researchers, funders, institutions, and people with Lived Experience. By aligning scientific rigour with ethical responsibility, digital phenotyping databanks can become transformative tools for advancing the global understanding and treatment of mental illness. NPP – Digital Psychiatry and Neuroscience; https://doi.org/10.1038/s44277-026-00066-z LAY SUMMARY Researchers are exploring how data from smartphones and wearables, such as movement, sleep, and phone use, might help detect and track depression and anxiety. This article outlines what is needed to build large, shared databanks of these data that are both useful and trustworthy, including better data standards, stronger privacy protections, greater participant control over data, and incentives for researchers to reuse data. International collaboration could accelerate the development of more effective ways to understand, detect and treat mental health problems across diverse communities.

  Building the foundations for global data banking in digital phenotyping for mental health · 2026 · DOI

• Artificially intelligent (AI) chatbots are increasingly used for mental health support, yet their safety guidelines and ethical structures remain unclear, particularly for youth populations.

  Debate: Conversational <scp>AI</scp> and young people's mental health – friend or foe? – the illusion of a safe space for youth mental health · 2026 · DOI

• Our results suggest that mobile chatbot applications may be promising tools to screen for depression indicators and inqui- ries during the perinatal and postpartum periods. However, further evaluation of this screening tool over time is needed to assess its accuracy in terms of indicators and diagnosis during follow-up. Findings contribute to evidence about usability and show promise for the efficacy of maternal health mobile applications in appropriate screening for maternal mental health indicators and inquiries. Findings also provide insight into closing the gap in maternal health disparities among Black mothers. This mobile application has the potential to be integrated as a tool within a compre- hensive care model for screening and early intervention dur- ing the perinatal and postpartum periods. Future research could continue to focus on these 2 areas: (1) evaluating the effectiveness of mobile application chatbots among larger samples and conducting assessments over time, and (2) iden- tifying risk factors and barriers to depression screening in perinatal and postpartum care. By continuing to explore these areas of research, the use of mobile applications like chatbots in maternal health can be further optimized and uti- lized in detecting maternal mental health conditions; and therefore, contribute to improved outcomes for Black moth- ers during the perinatal and postpartum periods.

  Using an AI-powered Mobile Application Chatbot to Address Maternal Depression Indicators and Inquiries in the Perinatal and Postpartum Periods: A Multimethod Analysis · 2026 · DOI

• Moreover, most prior trials have evaluated text-based chatbot interactions, whereas voice-based AI approaches capable of supporting spoken exchanges resembling clinician-patient conversational interactions during therapy remain understudied.

  A Phase-2 RCT of a Voice-based AI Coach for Depression and Anxiety · 2026 · DOI

• Eating disorders are among the most lethal psychiatric conditions, yet standardized benchmarks for evaluating AI conversations in this domain remain scarce, limiting rigorous cross-system comparison, harm potential, and measurement of improvement over time.

  mpathic Psychologist-led AI Clinical Tests Eating Disorders Benchmark (mPACT-ED-v1.0) · 2026 · DOI

• By combining real-time visual analysis with a supportive conver- sation, our system demonstrate the potential of AI to function as a low-barrier entry point for therapeutic art activity. At the same time, our expert evaluation warned several areas of improvement. Going forward, we will focus on refining the chatbot’s conversation flow to balance the depth and width of the interaction, integrat- ing visual interaction features that allow the AI to move beyond text and directly communicate with the user on canvas. We will develop a risk management module that monitors for signs of dis- tress, ensuring a safe and responsible wellbeing support. Prior to real-world user studies, these additions will need to be validated by more expert therapists with diverse training backgrounds and practicing experiences. Looking ahead, our research will transition from this initial ex- pert validation to a multi-phased user studies. We will begin with controlled lab studies to ensure the safety and usability of the sys- tem, and then proceed with longitudinal field studies to observe how this the system integrates into people’ daily lives. By gathering granular interaction patterns over time, we aim to understand how these multimodal interactions facilitated by AI shape users’ day-to- day emotional expression and self-reflection processes. Throughout the study, we will employ a proactive monitoring framework to identify and mitigate potentially triggering content, ensuring par- ticipant safety remains paramount. To provide a comprehensive evaluation, we will triangulate this qualitative interaction data with quantitative measures [2, 19], allowing us to measure the system’s long-term efficacy in fostering psychological resilience. DIS Companion ’26, June 13–17, 2026, Singapore, Singapore Lin et al.

  Exploring a Multimodal Chatbot as a Facilitator in Therapeutic Art Activity · 2026 · DOI

• We will exclude literature published in languages other than English and will search only 7 databases, which may limit some literature and add limitations to this study. Despite these limitations, this integrative review will provide implications for the development and implementation of chatbots for family caregivers of people with dementia.

  Evolution of Chatbots as an Educational and Supportive Digital Intervention for Family Caregivers of People With Dementia: Protocol for a Systematic Integrative Review · 2026 · DOI

• submitted then is the application’s effectiveness. During Similar to Germany, the PECAN procedure (35) allows for fast-track evaluation with early reimbursement by health insurers for one year, even without conclusive clinical evidence of the first year, manufacturers receive a monthly predefined compensation fee per patient. After one year, if the solution shows clinical benefits and is approved by the authorities to be a PECAN, the price can be for negotiated telemonitoring solutions. for DTx remain fixed but To request this early access, it is necessary to submit a simultaneous application to: 1. The Digital Health Agency (ANS) to obtain the solution’s certificate of conformity with the DMN interoperability and security framework; 2. The High Authority for Health (HAS) and the Ministers of Health and Social Security to assess its innovative nature, particularly in terms of clinical benefit or progress in the organization of care. PECAN process aims to complete evaluation and validation process in 90 days [60 days for the assessment by the Digital Health Agency (ANS) and the National Authority for Health (HAS) + 30 days for the final decision by the French Ministry of Health]. Upon receipt of favorable opinions, the decision on early access is published by decree of the Ministers of Health and Social Security within 30 days (36). Once the early coverage is established, there are 6 months to apply for the registration of the devices in the LPPR (List of Reimbursable Products and Services). The manufacturer must provide all the data required for a definitive evaluation of the device. The French government also pays close attention to the type of economic impact that the adoption of new digital therapies can have on the healthcare system: in particular, DTx manufacturers can now present, within the dossier, a budget impact analysis that illustrates the advantages brought by the digital therapy in terms of healthcare expenditure.

  The Italian landscape of digital therapeutics in a European context · 2026 · DOI

• Several limitations of this survey need to be considered. First, the overall response rate of the survey (5.8%) is quite low. However, a low response rate in itself does not necessarily indicate a biased sample.3 As the majority (65%) of survey respondents were APA members, a comparison of demographic characteristics between APA members versus past members or non-members within the sample was conducted (Table 3) to assess the potential difference between survey respondents and non- respondents. No statistical differences in terms of gender, race, ethnicity, sexual orientation, or disability status were found. However, the mean age of past members and non-members was found to be statistically younger than 2 American Psychological Association. (2016). 2015 survey of psychology health service providers. https://www.apa.org/workforce/publica- tions/15-health-service-providers 3 Czajka, J. L., & Beyler, A. (2016). Declining Response Rates in Federal Surveys: Trends and Implications. https://aspe.hhs.gov/system/files/ pdf/255531/Decliningresponserates.pdf AMERICAN PSYCHOLOGICAL ASSOCIATION 2 2021 SURVEY OF HEALTH SERVICE PSYCHOLOGISTS: TECHNICAL REPORT APA members (t = 3.18, p = 0.002). The fact that the majority of the sample are APA members and APA members tend to be older suggests the sample may be older than the health service psychologist workforce. As such, caution should be used when interpreting findings of this report. Second, the racial/ethnic groups used in the survey may mask the heterogeneity of populations within each category. Race categories of the survey were those included as APA’s data standard, which at the time did not include Arab, Middle Eastern or North African populations. The current APA data standard includes this population group and will be used in all future APA surveys. This addition would better reflect demographic diversity in the health service psychologist workforce. Finally, like all surveys, the current survey was subject to multiple sources of error, which were difficult to estimate and control, such as sampling error, coverage error, and error associated with non-response, question wording, and response options.

  Usability, Relevance, and User Engagement of a Government-Sponsored and Community Co-Designed Digital Mental Health Website During the COVID-19 Pandemic: A Pilot Study on Latino/a/x Adults · 2026 · DOI

• Adolescents increasingly use genAI tools in diverse ways (Madden et al., 2024), face unique risks in the absence of regulations and guidance (Microsoft, 2025), and will likely inherit an AI-dominated world without established best practices. These risks and opportunities require an urgent, coordinated response from researchers, policymakers, industry, and interventionists, grounded in developmental science and ethics.

  Coming of Age with Generative AI: The High Stakes of Automating Adolescent Development · 2026 · DOI

• Our future work will focus on integrating advanced clinical support systems to enhance reliability and provide a safety layer for high-risk cases, along with improved decision modules to better handle complex mental health scenarios. The system remains dependent on the accuracy of user-provided data, and the absence of direct clinical validation mechanisms is an important limitation.

  Swasthya Initiative - AI-Driven Student Mental Health Assessment · 2026 · DOI

• The present evaluation of the RELAX app demonstrates the initial results of a mHealth approach to the practical implementation of the JITAI concept in the domain of occupational stress management. The application’s objective is to deliver interventions that are not only time efficient but also content focused, with the aim of optimally addressing the individual’s situation and needs. This approach builds on https://mhealth.jmir.org/2026/1/e79642 The assessment of multiple levels of stress markers made it possible to compare them, whereas previous studies of comparable approaches have focused on only a few parameters [13,23]. This approach allowed for the identification of variability in the effects across dimensions, which, although not in the expected manner, may provide valuable insights from alternative perspectives. It is acknowledged that stress markers do not always align, particularly between physiological and subjective parameters. The findings from this study lay the foundation for future research to dive deeper into this divergence and potentially link it to underlying factors within the specific mHealth context. This study fulfills the quality requirements of a pilot study. However, it must therefore also be evaluated with the limited significance of such a study. The absence of a control group without intervention, as previously identified as a limitation in the study protocol, prevents clear attribution of all changes in stress parameters due to the intervention. The study design used here does not provide the internal validity required for statements on the effectiveness of the intervention. At this juncture, the outcomes of this study can only be interpreted as effects that occurred during the study period, which may provide an indication of effectiveness that must be validated with a controlled follow-up study. Furthermore, as previously stated in the discussion, the study period was relatively brief, thereby limiting the opportunities for user intervention. The small sample size precluded the ability to draw definitive conclusions; thus, the findings should be considered exploratory in nature. This is particularly evident in the group randomization, where the sample sizes were further diminished. Following the implementation of technical stabilization updates to the application and the completion of its development phase, which will enable fully automated intervention selection, the evaluation should be expanded to include a more substantial sample size and an extended study duration. Moreover, the employees’ daily experiences in their highly individual roles and in different industries are likely to have generated such a variable set of situations that the stress events within and between individuals call into question the comparability of the EMA data. The app’s capacity to adapt to each unique situation and individual user could be considered overly ambitious for the current stage of development, particularly in light of the absence of complete JMIR Mhealth Uhealth 2026 | vol. 14 | e79642 | p.

  Multimodal Personalized Mobile Health Just-in-Time Adaptive Intervention for Occupational Stress Management: Pilot Study · 2026 · DOI

• Future research could address this limitation through longitudinal designs or quasi experimental approaches that allow for the evaluation of changes in well being over time, including pre and post intervention assessments. This suggests that the benefits associated with access to reliable information and personalised digital support may extend across diverse migrant experiences, rather than being limited to a specific subgroup of users.

  Promoting inclusion through information: a study on the well-being of migrant women using a mobile application in Chile · 2026 · DOI

• A u t h o r M a n u s c r i p t A u t h o r M a n u s c r i p t A u t h o r M a n u s c r i p t A u t h o r M a n u s c r i p t Horwitz et al. Page 8 et al., 2025). Our use of ‘number of days used’ was influenced by both theoretical and practical considerations, as inconsistencies emerged based on device and application when examining engagement at the minutes-level—while number of days used was the most reliable metric, it was unable to distinguish longer and shorter durations of use, or separate episodes within a given day, and does not allow for insights into average length of use or concrete recommendations related to appropriate “dosage” of a DMHI to expect an effect. Furthermore, our measurement of engagement was based strictly on behavioral usage patterns of the app, which has the potential to overlook integration of app content (e.g., CBT skills learned from modules) in offline, real-world contexts, and does not capture potential cognitive/affective aspects of engagement (e.g., how closely they were attending to the content while in the app). Finally, our clinical metrics were static measures of baseline clinical severity and our analyses could not account for the influence of dynamic symptoms changes on app usage over the course of the study period.

  Sociodemographic and clinical predictors of digital mental health intervention engagement among treatment-seeking psychiatric outpatients · 2026 · DOI

• Online CBT, digital self-management Reduced anxiety & depression AI dialogue model, ontology graph, BERTbased extraction, hybrid intent recognition High personalization…

  A HYBRID HUMAN-AI WEB SYSTEM FOR REAL-TIME MENTAL HEALTH COUNSELLING AND CHRONIC DISEASE MANAGEMENT · 2026 · DOI

• The responsible integration of AI into college students’ mental health education requires coordinated action at multiple levels. Because field-specific intervention evidence remains limited in several areas, these recommendations draw on both higher education research and cautiously interpreted insights from adjacent fields, particularly healthcare AI, AI ethics, and digital inequality research. They are therefore intended as proportionate guidance for implementation rather than as claims that every recommendation has already been directly validated in college mental health education settings.

  Artificial intelligence in college students’ mental health education: opportunities, challenges, and strategic responses · 2026 · DOI

• for This paper presented the MCI Cognitive Care App, training an AI-powered personalized cognitive platform individuals with Mild Cognitive Impairment. By integrating reinforcement learning, gamification, and collaborative care, the system addresses key limitations of traditional cognitive rehabilitation tools. Future work will focus on largescale evaluations, expansion of cognitive exercises, integration of wearable data, and exploration of advanced personalization models. REFERENCES S. Belleville, S. Gilbert, F. Fontaine, S. Gagnon, D. Menard, and L.´ Gauthier, “Improvement of episodic memory in persons with mild cognitive impairment and healthy older adults: Evidence from a cognitive intervention program,” Neuro psychologia, vol. 44, no. 12, pp. 2161–2170, 2007. and A. Bahar-Fuchs, M. Clare, and L.

  MCI Cognitive Care App: An AI-Powered Personalized Platform for Cognitive Training in Mild Cognitive Impairment · 2026 · DOI

• • The chosen ROC thresholds were not manually or externally checked and were data based. Thus, other choices of cut-off Frontiers in Artificial Intelligence 13 frontiersin.org Dessai et al. 10.3389/frai.2026.1769286 can have an impact on the sensitivity-specificity trade-off and change the results of classification. • Another part of this work in the future is systematic assessment and enhancement of the RoBERTa model to be less susceptible to sarcastic, ambiguous, and culturally specific responses with the help of specific test sets and human-in-the- loop verification. • Since the model was trained using data that was synthetically generated, its performance might not be able to reflect the range of human conversations in the real world. • The current analysis is only on the population of the students in the MHP data. In this regard, the results can not necessarily be extrapolated to other population segments. The framework, however, is flexible and can be reconfigured with population-specific data and ROC thresholds to have a greater applicability. • Although the proposed model achieved nearly perfect precision and specificity on the synthetic dataset, such performance is influenced by the structured and rule-based data generation process. Real-world datasets typically contain greater variability, ambiguity, and noise. Future work will involve evaluating the model on independently collected real- world data to further assess robustness and generalizability.

  Adaptive emotion-aware chatbot for mental health diagnosis using recurrent reinforcement learning and transformer models · 2026 · DOI

• • • • • • The following recommendations were drawn from the summary of findings and conclusions: Incorporate an upgraded AI model to improve response accuracy. Utilize a machine learning model with a blank slate and continuously feed data; however, this requires a lot of preloaded data and takes time. Customize the outcome so the counselor can easily access it and provide a summary of each user's complaints. To find and fix any vulnerabilities, do security audits on a regular basis. Give end users thorough instruction on data security and privacy procedures. Conduct internal audits on a regular basis to make sure ISO 25010 requirements are being followed. Create a documentation procedure to monitor and prove compliance with each quality standard.

  Open Artificial Intelligence Integration Platform for Guidance Counselor Monitoring of the Driven Mental Health Support System of Students · 2026 · DOI

• The three human-centered perspectives applied to this CDSS work (distributed cognition, situated learning, and infrastructural inversion) are not comprehensive for understanding psychotherapeutic contexts. Alternative theoretical lenses and their utility for designing and evaluating clinical decision support systems in intensive outpatient PTSD care have not been identified or tested.

  Human-centered Perspectives on a Clinical Decision Support System for Intensive Outpatient Veteran PTSD Care · 2026 · DOI

Most-cited papers in Digital Mental Health Interventions

• Effectiveness of online mindfulness-based interventions in improving mental health: A review and meta-analysis of randomised controlled trials · Clinical Psychology Review · 2016 · 891 citations
• Digital interventions for the treatment of depression: A meta-analytic review. · Psychological Bulletin · 2021 · 389 citations
• Predictors of treatment dropout in self-guided web-based interventions for depression: an ‘individual patient data’ meta-analysis · Psychological Medicine · 2015 · 382 citations
• Large language models could change the future of behavioral healthcare: a proposal for responsible development and evaluation · npj Mental Health Research · 2024 · 283 citations
• Too human and not human enough: A grounded theory analysis of mental health harms from emotional dependence on the social chatbot Replika · New Media & Society · 2022 · 277 citations
• Internet-delivered cognitive behavior therapy for children and adolescents: A systematic review and meta-analysis · Clinical Psychology Review · 2016 · 274 citations
• Internet-based cognitive behavioural therapy for subthreshold depression in people over 50 years old: a randomized controlled clinical trial · Psychological Medicine · 2007 · 261 citations
• Preliminary Evaluation of PTSD Coach, a Smartphone App for Post-Traumatic Stress Symptoms · Military Medicine · 2014 · 230 citations
• Artificial Intelligence and Chatbots in Psychiatry · Psychiatric Quarterly · 2022 · 222 citations
• Current evidence on the efficacy of mental health smartphone apps for symptoms of depression and anxiety. A meta‐analysis of 176 randomized controlled trials · World Psychiatry · 2024 · 189 citations

Most recent work

• Potentially Harmful Consequences of Artificial Intelligence ( <scp>AI</scp> ) Chatbot Use Among Patients With Mental Illness: Early Data From a Large Psychiatric Service System · Acta Psychiatrica Scandinavica · 2026
• Using Wearables in Mental Health Care for Children and Adolescents: A Scoping Review · Research on Child and Adolescent Psychopathology · 2026
• Using an AI-powered Mobile Application Chatbot to Address Maternal Depression Indicators and Inquiries in the Perinatal and Postpartum Periods: A Multimethod Analysis · INQUIRY The Journal of Health Care Organization Provision and Financing · 2026
• TherapyProbe: Generating Design Knowledge for Relational Safety in Mental Health Chatbots Through Adversarial Simulation · 2026
• Human-centered Perspectives on a Clinical Decision Support System for Intensive Outpatient Veteran PTSD Care · 2026
• Breaking Negative Cycles: A Reflection-to-Action System for Adaptive Change · 2026
• Exploring Students’ Perceptions and Usage of Artificial Intelligence in Supporting Mental Health: A Preliminary Study in Higher Education in Qatar · Healthcare · 2026
• Toward Participatory Precision Health With Co-Designed Recommendations: Systematic Review of Just-in-Time Adaptive Interventions in Adolescents and Young Adults · Journal of Medical Internet Research · 2026
• Individual-Level Modeling of Depressive Symptom Severity Using Smartphone and Wearable Data: A Time-Aware 1-Year Study with Feature-Group Contributions (Preprint) · 2026
• Generative artificial intelligence in mental health: A preliminary study on automating materials development for cognitive bias modification · International Journal of Mental Health · 2026

Related topics in Psychology