[Explainable AI Team Seminar] NLP for Mental Health: Risk, Promise, Responsibility by Prof. Irina Gurevych

RIKEN AIP's Explainable AI Team hosts an online seminar by Iryna Gurevych, Professor of Ubiquitous Knowledge Processing at the Technical University of Darmstadt and a past president of the Association for Computational Linguistics. The session runs on Zoom from 14:30 to 16:00 JST, and the joining link is sent only to registered participants. Registration is open to anyone. The talk argues that natural language processing can genuinely help with mental health care, but that progress depends on clinical grounding, real data, and responsible evaluation rather than simply larger models. Drawing on work from her group, Prof. Gurevych shows how large language models can turn therapy transcripts into treatment-relevant structure, how clinically grounded synthetic data eases the privacy bottleneck the field faces, and why evaluation should reflect what clinicians actually care about instead of what is easy to benchmark. She closes with a case for building mental health AI that is personalized, privacy-aware, and responsibly evaluated: tools meant to support clinicians rather than replace them. The seminar suits NLP and machine learning researchers, clinical informatics practitioners, and anyone working on the evaluation and safety of language models in high-stakes settings.