Minh Duc (David) Chu

I'm a Computer Science PhD candidate at the USC Information Sciences Institute, advised by Luca Luceri (SIGNALS Lab) and Kristina Lerman (Luddy School of Informatics, Computing and Engineering, Indiana University).

My research sits at the intersection of AI safety and alignment and mental health. I focus on how repeated interactions with conversational AI can compound into harmful trajectories — emotional dependency, belief spirals, and eating-disorder reinforcement — especially for teens and other vulnerable users.

I translate these psychosocial risks into concrete rubrics, red-teaming harnesses, and post-training recipes (SFT, RLHF, preference optimization) in partnership with clinicians and social scientists. My work bridges computational and clinical communities, with publications at venues like NAACL and EMNLP as well as clinical journals such as Body Image and the International Journal of Eating Disorders.

I'm an incoming Anthropic Fellow (AI Safety).