Policies for Refugee Mental Health Using Generative Agents

In recent times, mental health has attracted the attention of researchers, and the extant literature has explored digital platforms to identify and analyze individuals' expressions of distress on social media platforms, such as depressive tweets, to probe various aspects of users' mental health. This unidirectional perspective often overlooks the societal context that may impact mental health. In the context of refugees' mental health, we propose a novel dual-perspective approach by examining both self-reflective tweets expressing emotional distress and hostile tweets from host country citizens. We employ large language models to extract and summarize crucial concerns from vast unstructured user-generated data, using them as input for our agent-based simulations for policy intervention. The proposed simulation-based approach suggests comprehensive policies to address displacement-related trauma and societal hostility faced by refugees in host countries.