Yaman Yu's Preliminary Exam
PhD student Yaman Yu will present her dissertation proposal, "Safeguarding and Empowering Youth in the Era of Generative AI." Her preliminary examination committee includes Professor Yang Wang (chair); Assistant Professor Haohan Wang; Associate Professor Yun Huang; and Associate Professor Sauvik Das, Human-Computer Interaction Institute, Carnegie Mellon University.
Abstract
The rapid advancement of generative artificial intelligence is transforming how individuals communicate, learn, and participate in digital life. Adolescents and young adults have consistently been early adopters of emerging technologies and are now highly engaged users of generative AI platforms, such as conversational agents and Generative AI-augmented tools. However, their developmental stage, evolving sense of identity, and limited ability to critically evaluate AI-generated content make them especially vulnerable to a range of emerging risks. These risks are often subtle, socially embedded, and difficult to detect using existing moderation systems that are not designed with youth-specific needs in mind.
This growing disconnect between youth engagement and the safeguards currently in place presents both a technical and ethical challenge. Without proactive research and intervention, generative AI systems may reinforce harmful behaviors, normalize inappropriate advice, or introduce long-term psychological and social harms. Current platform designs often fail to account for the ways in which youth interpret, trust, and act upon AI outputs, leading to missed opportunities for meaningful protection and support.
In my proposal, I aim to address this gap through a mixed-methods investigation into how youth interact with generative AI and what risks emerge in these interactions. It draws on empirical studies involving both teenagers and parents to identify patterns of use, perceived threats, and limitations of existing safety mechanisms. Building on these findings, I aim to provide a multi-layered taxonomy of youth-specific risks, a benchmark to evaluate safeguard performance, and the design of targeted intervention tools that support real-time risk detection and age-appropriate responses. These contributions provide a foundation for building safer, more inclusive AI systems that align with the needs and expectations of young users, and offer practical guidance for platform developers, educators, and policymakers.
Questions? Contact Yaman Yu.