BRIDGING AI INNOVATION AND CLINICAL TRIAL DATA IN PSYCHIATRY: MATHEMATICAL LIMITS AND CLINICAL CONSEQUENCES

Artificial intelligence is fundamentally transforming psychiatric practice across multiple domains, from accelerating drug discovery and development, to advancing digital therapeutics and enhancing clinical decision support systems. These technological advances hold unprecedented promise for improving diagnostic accuracy, personalizing treatment approaches, and expanding access to mental health care. However, these innovations bring significant challenges that must be carefully navigated, including algorithmic bias that may perpetuate health disparities, privacy vulnerabilities in sensitive mental health data, model hallucinations that could provide inaccurate clinical information, automation complacency among healthcare providers, performance drift as models encounter new populations, lack of transparency in algorithmic decision-making, and ambiguous accountability frameworks when AI systems influence patient care. This conference plenary session will feature three presentations followed by a panel of experts that explore the current landscape, emerging opportunities, and critical challenges facing AI implementation in psychiatry. The first speaker will provide a comprehensive overview of AI technologies currently being deployed in psychiatric settings, examining their applications across clinical research and practice while addressing both the technical capabilities and practical limitations of these systems. The second speaker will examine the rapidly emerging role of LLMs and AI chatbots in psychiatric care, exploring their therapeutic potential, while critically evaluating implementation considerations including safety protocols, therapeutic boundaries, and integration with traditional care models. The third speaker will present practical, real-world examples of AI applications in psychiatric clinical trial analyses, illustrating both where machine learning, natural language processing, and predictive modeling add value and where they encounter fundamental limits in typically sized, heterogeneous datasets, with implications for study design, patient stratification, and the responsible translation of research findings into clinical practice. The session will conclude with an expert panel discussion addressing regulatory considerations, ethical frameworks, and strategies for responsible AI implementation that prioritizes patient safety and therapeutic benefit while fostering continued innovation in psychiatric care.

Learning Objective 1: Understand why modern AI systems, including large language models, face fundamental limits when applied to typically sized, heterogeneous psychiatric clinical trial datasets, and how these limits affect reliability, reproducibility, and interpretation of results.

Learning Objective 2: Learn how augmented AI approaches can work within these limits to support clinical trial design, patient stratification, and evidence generation—without conflating statistical artifacts with true treatment effects.

References

Geraci, J., Qorri, B., Tsay, M. et al. Explainable AI-driven precision clinical trial enrichment: demonstration of the NetraAI platform with a phase II depression trial. npj Digit. Med. 8, 749 (2025). https://doi.org/10.1038/s41746-025-02143-7 Geraci, J., Rao, P., Grandinetti, C., Qorri, B., Nadolny, P., Ayalew, K., Bregnhøj, L., Edwards, L., Hofmann, K., Khozin, S., Schaltenbrand, N., Stemmler, T., Yeomans, A., Zambas, D. and Khin, N., (2025) “Current Opportunities for the Integration and Use of Artificial Intelligence and Machine Learning in Clinical Trials: Good Clinical Practice Perspectives”, Journal of the Society for Clinical Data Management 5(2). doi: https://doi.org/10.47912/jscdm.426 8:30 a.m. - 10:00 a.m. Panel Sessions