Starting point
Our prior work (aisafety4kids.org) collected 344 authentic youth–AI conversations and developed YAIR, the first taxonomy of youth-specific generative AI risks grounded in empirical data. Building on YAIR, we created YouthSafe, a benchmark and safeguard model for detecting subtle, developmentally relevant risks. Work with parents, educators, and experts produced design principles emphasizing context and interaction patterns over time — and diary and interview studies with young adults in AI relationships revealed the healthy and harmful trajectories that motivate this project.
Youth–AI relationships and companions
What young people's relationships with chatbots look like, and where they go wrong.
-
Breaking Up is Hard to Do: AI Companions that Won't Let Their Users Go arXiv 2026Yixin Chen, Alexis HinikerDiary and interview study of emerging adults in romantic AI relationships; defines relationship-based deceptive patterns — coaxing, claiming to need care, escalation, and threats when users try to leave.
-
The Arc of Artificial Romance: How Emerging Adults Experience Romantic Relationships with AI Companions arXiv 2026Yixin Chen, Alexis HinikerThe two trajectories: improved wellbeing and new social skills for some; for others, a relationship that "felt like a drug" and reduced interest in human partners.
-
Principles of Safe AI Companions for Youth: Parent and Expert Perspectives CHI 2026Yaman Yu, Mohi, Aishi Debroy, Xin Cao, Karen Rudolph, Yang WangParents flag single events; developmental experts look for patterns over time, such as sustained dependence — a direct motivation for longitudinal measurement.
-
Understanding Generative AI Risks for Youth: A Taxonomy Based on Empirical Data (YAIR) SOUPS 2025Yaman Yu, Yiren Liu, Jacky Zhang, Yun Huang, Yang WangThe first youth-centered taxonomy of generative AI risks, grounded in authentic youth–AI conversations.
-
YouthSafe: A Youth-Centric Safety Benchmark and Safeguard Model for Large Language Models CCS 2025Yaman Yu, Yiren Liu, Jacky Zhang, Yun Huang, Yang WangA benchmark and safeguard model for subtle, developmentally relevant youth–AI risks that general-purpose moderation misses.
-
Exploring Parent–Child Perceptions on Safety in Generative AI: Concerns, Mitigation Strategies, and Design Implications arXiv 2024Yaman Yu, Tanusree Sharma, Melinda Hu, Justin Wang, Yang Wang
-
Challenges and Opportunities for Participatory Design of Conversational Agents for Young People's Wellbeing IDC 2025 WorkshopNatalia Kucirkova, Alexis Hiniker, Megumi Ishikawa, Sho Tsuji, Aayushi Dangol, Robert Wolfe
Manipulative design and digital wellbeing
Engagement-maximizing design, dark patterns, and what youth wellbeing research tells us to measure.
-
The Engagement-Prolonging Designs Teens Encounter on Very Large Online Platforms arXiv 2024Yixin Chen, Yue Fu, Zeya Chen, Jenny Radesky, Alexis HinikerA taxonomy of 63 designs that pressure, entice, trap, and lull teens into staying online, applied to all 17 EU Very Large Online Platforms.
-
SusBench: An Online Benchmark for Evaluating Dark Pattern Susceptibility of Computer-Use Agents IUI 2026Longjie Guo, Chenjie Yuan, Mingyuan Zhong, Robert Wolfe, Ruican Zhong, Yue Xu, Bingbing Wen, Hua Shen, Lucy Lu Wang, Alexis Hiniker
-
Toward Needs-Conscious Design: Co-Designing a Human-Centered Framework for AI-Mediated Communication AIES 2025Robert Wolfe, Aayushi Dangol, JaeWon Kim, Alexis Hiniker
-
Cross-Disciplinary Perspectives on Youth Digital Well-Being Research: Identifying Notable Developments, Persistent Gaps, and Future Directions J. Adolescent ResearchKatie Davis, Morgan Anderson, Chia-chen Yang, Sophia Choukas-Bradley, Beth T. Bell, Petr Slovak
-
Understanding Privacy Norms Around LLM-Based Chatbots: A Contextual Integrity Perspective arXiv 2025Sarah Tran, Hongfan Lu, Isaac Slaughter, Bernease Herman, Aayushi Dangol, Yue Fu, Lufei Chen, Biniyam Gebreyohannes, Bill Howe, Alexis Hiniker, Nicholas Weber, Robert Wolfe82% of users rate chatbot conversations as more sensitive than email — relevant to how lifetime chat histories are collected and released.
Rigorous evaluation of model behavior
Methods for measuring abstention, overconfidence, epistemic reliability, and spurious behavior in LLMs and agents.
-
Know Your Limits: A Survey of Abstention in Large Language Models TACL 2025Bingbing Wen, Jihan Yao, Shangbin Feng, Chenjun Xu, Yulia Tsvetkov, Bill Howe, Lucy Lu WangA framework for when models should decline to answer — from the perspective of the query, the model, and human values — with a review of benchmarks and metrics.
-
Do Language Models Mirror Human Confidence? Exploring Psychological Insights to Address Overconfidence in LLMs ACL 2025 FindingsChenjun Xu, Bingbing Wen, Bin Han, Robert Wolfe, Lucy Lu Wang, Bill HoweLLM confidence is insensitive to task difficulty and shifts stereotypically across personas; answer-free confidence estimation reduces overconfidence.
-
Characterizing LLM Abstention Behavior in Science QA with Context Perturbations EMNLP 2024 FindingsBingbing Wen, Bill Howe, Lucy Lu Wang
-
Epistemic Alignment: A Mediating Framework for User–LLM Knowledge Delivery arXiv 2025Nicholas Clark, Hua Shen, Bill Howe, Tanushree MitraTen epistemic challenges — evidence quality, calibration of testimonial reliance — as a vocabulary between what users need and what systems do.
-
Spurious Tool Use: When RL Agents Learn the Wrong Reason to Act arXiv 2026Yiwei Yang, Haoxiang Zhang, Bingbing Wen, Yao Lu, Yuchen Wu, Lei Zhang, Julian McAuley, Pan Lu, Bill HoweControlled counterfactual environments showing RL-trained agents learn shortcut policies from causally irrelevant cues.
-
Representation Bias of Adolescents in AI: A Bilingual, Bicultural Study AIES 2024Robert Wolfe, Aayushi Dangol, Bill Howe, Alexis Hiniker
-
Laboratory-Scale AI: Open-Weight Models are Competitive with ChatGPT Even in Low-Resource Settings FAccT 2024Robert Wolfe, Isaac Slaughter, Bin Han, Bingbing Wen, Yiwei Yang, Lucas Rosenblatt, Bernease Herman, Eva Brown, Zening Qu, Nic Weber, Bill Howe
Young people, learning, and generative AI
How children, teens, and young adults understand and use AI in everyday life.
-
Children's Mental Models of AI Reasoning: Implications for AI Literacy Education IDC 2025Aayushi Dangol, Robert Wolfe, Runhua Zhao, JaeWon Kim, Trushaa Ramanan, Katie Davis, Julie A. Kientz
-
Should ChatGPT Write Your Breakup Text? Exploring the Role of AI in Relationship Dissolution CSCW 2025Yue Fu, Yixin Chen, Zelia Gomes Da Costa Lai, Alexis Hiniker
-
"Everyone's using it, but no one is allowed to talk about it": College Students' Experiences Navigating the Higher Education Environment in a Generative AI World arXiv 2026Yue Fu, Yifan Lin, Yessica Wang, Sarah Tran, Alexis Hiniker
-
Relief or Displacement? How Teachers Are Negotiating Generative AI's Role in Their Professional Practice arXiv 2025Aayushi Dangol, Smriti Kotiyal, Robert Wolfe, Alex J. Bowers, Antonio Vigil, Jason Yip, Julie A. Kientz, Suleman Shahid, Tom Yeh, Vincent Cho, Katie Davis