This project examines how teens interact in real time with AI chatbots that use either a relational style (warm, affiliative, commitment-oriented language) or a transparent style (explicit nonhuman framing and informational tone). We bring adolescents into the lab to complete structured conversations with both chatbot styles, allowing us to observe interaction patterns, preferences, and moment-to-moment responses beyond vignette-based judgments. To understand for whom these designs may be most helpful or most risky we recruit youth across a range of social stress experiences and compare how conversational style shapes outcomes such as perceived support, trust, anthropomorphism, and potential emotional over-reliance. The study also includes functional near-infrared spectroscopy (fNIRS) to measure prefrontal brain activation during the conversations, helping us examine underlying social-cognitive and emotion-regulation processes as teens engage with relational versus transparent chatbots. Findings will inform developmentally grounded guidance for youth-safe chatbot design and practical recommendations for families, educators, and policymakers.

‍