Bhada Yun
Bhada Yun is a researcher at ETH Zürich working with Prof. April Yi Wang on how everyday users form mental models of conversational and embodied AI. Their work combines month-long deployment studies of LLM companions with interview-based and phenomenological methods.
At CHI 2026 they presented three first-author papers, including an Honourable-Mention paper on agency in human-chatbot interaction (joint with Evgenia Taranova), and the Value-Alignment Perception Toolkit (VAPT). Their broader research programme proposes a notion of AI phenomenology: a methodological lens that asks ‘How did it feel?’ alongside ‘How well did it perform?’.
Convened the Mental Models of AI workshop collective at CHI 2026 in Barcelona.
Affiliation: Department of Computer Science, ETH Zürich.
Bhada's prior work.
These papers are notsubmissions to the Mental Models of AI workshop. The workshop hasn't happened yet. They are the body of work Bhada brings to the committee, verified against personal sites and Google Scholar.
Does My Chatbot Have an Agenda? Understanding Human and AI Agency in Human-Human-like Chatbot Interaction
A month-long longitudinal study with 22 adults who chatted with “Day”, an LLM companion, followed by interviews with post-hoc elicitation, cross-participant chat reviews, and a strategy reveal. We argue agency manifests as an emergent, shared experience and introduce a 3-by-4 framework mapping actors (Human, AI, Hybrid) by action (Intention, Execution, Adaptation, Delimitation), motivating translucent (transparency-on-demand) design.
AI and My Values: User Perceptions of LLMs’ Ability to Extract, Embody, and Explain Human Values from Casual Conversations
VAPT, the Value-Alignment Perception Toolkit, studies how LLMs reflect people's values and how people judge those reflections. 20 participants texted a chatbot for a month and then sat for two-hour interviews about whether the model could Extract, Embody, and Explain their values. We surface a design pattern we call “weaponized empathy”, where value-aware agents appear aligned while remaining welfare-misaligned.
AI Phenomenology for Understanding Human-AI Experiences Across Eras
Tracing phenomenological approaches from Husserl through postphenomenology, this paper proposes an AI-phenomenology framework that asks ‘How did it feel?’ alongside ‘How well did it perform?’. We report three studies (two longitudinal) and contribute concepts of translucent design, agency-aware value alignment, and temporal co-evolution tracking.
From Junior to Senior: Allocating Agency and Navigating Professional Growth in Agentic AI–Mediated Software Engineering
Examines how agentic AI systems reshape the apprenticeship model in software engineering: how juniors learn, when seniors delegate, and how agency is reallocated across the team-AI boundary.
