When AI Becomes Relational: Trust, Intimacy and Responsibility
As AI becomes more conversational, people are beginning to interact with it in new ways. Some use it to brainstorm ideas, others to learn, seek advice, or simply have someone to talk to. As these interactions become more personal, they also raise new questions.
What responsibilities do developers have when people begin to experience these interactions as emotionally meaningful? How should we think about trust, memory, consent, emotional safety, or vulnerable users? And are today's ways of evaluating AI enough when the quality of the relationship starts to matter as much as the quality of the response?
Join us for an Off-Script panel exploring these questions through the perspectives of researchers, builders, and governance practitioners. Rather than debating whether relational AI is good or bad, we'll examine the opportunities, risks, and design considerations that come with building AI systems people increasingly interact with on a human level.
More About the Panelists
Mabel Loh is the founder of Maibel, a Singapore-based AI startup building emotionally intelligent wellness companions for women. Combining behavioral science, narrative design and conversational AI, Maibel explores how technology can support sustainable wellness habits through ongoing, emotionally meaningful interactions.
Mabel’s work focuses on relational AI safety, behavioral evaluation, memory and user trust. Drawing on her experience building relational AI products for users across Southeast Asia, she brings a practical perspective on designing systems that remain safe, consistent and genuinely useful over time.
Hakim Norhashim (Research Associate, NUS AI Institute)
Hakim Norhashim is a Research Associate at the NUS AI Institute, where his work centres on AI governance and alignment. His research investigates how large language models reflect and engage with local cultural contexts, developing frameworks for cultural alignment that guide AI governance practices. He focuses on how such practices can be operationalised through both human-centred and technical measures, while accounting for the social contexts in which AI systems operate. His paper, Measuring Human-AI Value Alignment in Large Language Models, was published in the proceedings of the Seventh AAAI/ACM Conference on AI, Ethics, and Society.
Dr. Gerard Yeo's research sits at the intersection of AI and psychology, where he studies how AI and large language models (LLMs) reason about human beliefs, emotions, and social interactions, and how these capabilities shape the way people perceive and trust AI systems. More recently, his work has focused on evaluating the trustworthiness of LLMs and developing approaches to support the safe and responsible governance of AI. He received his Ph.D. in Data Science from the National University of Singapore (NUS), where he also earned his Bachelor's and Master's degrees in Psychology. By combining insights from AI, psychology, and computational social science, his research seeks to better understand the relationship between people and AI, and the implications for the design, evaluation, and governance of AI systems.
More About the Moderator
Dr Kenny Choo (Assistant Professor, Singapore University of Technology and Design (SUTD))
Dr Kenny Choo directs the Context-Aware Interaction Lab (CAILab) and researches how people build trust in, rely on, and negotiate decision-making authority with AI systems. His work spans HCI's top venues: CHI (two 2026 papers, both earning Honourable Mention Awards — top 5% of submissions — including one on AI agents as personal advocates in advance care planning), CSCW, and ACM Transactions on Computer-Human Interaction (TOCHI), where he published CoAIcoder, on AI-mediated collaborative qualitative analysis. He also publishes across NLP (ACL, EMNLP), ubiquitous computing (MobiSys, IEEE Pervasive Computing), and AI (IJCAI), reflecting a research programme that bridges human-centred design with technical AI/ML systems. He brings this cross-disciplinary lens to moderating this panel on trust, intimacy, and responsibility in relational AI.
More About the Series
Off‑Script is a conversation series by Lorong AI that brings together practitioners to explore ideas at the intersection of technology, design, and society. Through candid dialogue, live feedback, demos, and shared experiences, the series creates space for thoughtful exchange beyond slide decks.
Get involved: Lorong AI | Speaker Sign-up | WhatsApp Community | LinkedIn | X