JMIR Publications

ABSTRACT

Drug addiction is a growing global public health crisis that affects not only individuals but entire family systems. While effective therapeutic approaches exist, many families struggle to access qualified addiction therapists or receive timely, practical guidance on how to support a loved one with substance use disorder. Moreover, families often face a painful emotional dilemma: enabling addictive behavior versus enforcing boundaries that may lead to homelessness, incarceration, or further harm. This paper proposes a novel, complementary approach: the development of a language model fine-tuned on recovery-oriented data, including guidance from addiction therapists and lived experience narratives from recovering addicts. The goal is not to replace professional treatment, but to provide accessible, empathetic, yet emotionally neutral and experience-informed advisory support for parents and loved ones. We argue that recent advances in LLMs, fine-tuning techniques, and alignment methods make this approach both feasible and timely. We outline the motivation, conceptual foundations, technical feasibility, and potential societal impact of such a system.