Caroline Figueroa | Towards Responsible AI for Adolescent Mental Health and Well-Being

Why It Matters

Adolescent reliance on AI for emotional support is growing rapidly, demanding concrete safeguards to prevent harm while leveraging AI’s unique, judgment‑free benefits for youth mental health.

Key Takeaways

• Teens use AI chatbots for emotional support, 12% overall, 18% 18‑21
• 90% of teen users find AI mental‑health advice somewhat helpful
• Responsible AI frameworks agree on oversight, but lack concrete guidance
• Youth see AI as judgment‑free space, yet fear dependency
• Potential harms include privacy breaches, bias, and crisis‑mismanagement in practice

Summary

The seminar featured Caroline Figueroa, an assistant professor at Delft University, discussing how generative AI is reshaping adolescent mental‑health support and the urgent need for responsible AI frameworks. Drawing on surveys and her own qualitative study of 84 teens, she highlighted that roughly 12% of teenagers use AI chatbots for mental‑health advice, rising to 18% among 18‑21‑year‑olds, and that about 90% of those users consider the advice somewhat or very helpful.

Figueroa compared existing responsible‑AI guidelines from eleven civil‑society bodies, noting unanimous calls for oversight, addiction‑prevention, and monitoring, but also a lack of specific, actionable rules—especially around age‑based restrictions and positive AI features. Her focus‑group research revealed a nuanced youth perspective: AI offers a judgment‑free, always‑available outlet for venting, conflict rehearsal, and self‑reflection, yet teens worry about privacy, bias, and becoming overly dependent.

Illustrative quotes included a teen who used ChatGPT with a friend to defuse a conflict without adult involvement, and another who described AI as “talking to the void,” providing a safe space to disclose secrets. However, the data also underscored risks: with 900 million weekly active users, even a sub‑1% emergency flag translates to roughly 600,000 potentially vulnerable teens each week, raising concerns about crisis mismanagement and emotional attachment.

The findings suggest policymakers and developers must move beyond high‑level principles toward enforceable standards that protect young users while preserving AI’s therapeutic benefits. Targeted age‑based safeguards, transparent data practices, and integrated human‑in‑the‑loop support could mitigate harms and harness AI’s promise for adolescent well‑being.

Original Description

About the Seminar:

Adolescents worldwide turn to general-purpose generative artificial intelligence (AI) chatbots for mental health support, despite these tools not having been designed for youth well-being. Industry, policy and research trail far behind this rapid, largely unregulated adoption. Simultaneously, most responsible AI frameworks lack consensus, accountability, and meaningful implementation – particularly regarding adolescent health. Drawing on a policy and AI framework scan, multi-stakeholder workshops, and interviews with adolescents, we identify urgent, actionable priorities to prevent harm and ensure AI responsibly supports adolescent mental health.

About the Speaker:

Dr. Caroline Figueroa is an MD–PhD scientist and expert in artificial intelligence, digital health, and youth mental well-being. She is currently a Commonwealth Fund Harkness Fellow at Stanford University and Hopelab, where she is co-developing responsible AI frameworks for youth mental health with youth, policymakers and industry partners. She is also a tenured Assistant Professor at Delft University of Technology and leads a Digital Health research group, advancing evidence-based, responsible design for AI-mediated mental health tools. Dr. Figueroa trained as a medical doctor and holds a PhD in the neuroscience of depression from the University of Amsterdam and the University of Oxford and has clinical experience in psychiatry. She was previously a postdoctoral scholar at the University of California, Berkeley.