Bhada Yun

Bhada Yun

Current Institution
ETH Zürich
Current Research Positions
Researcher at IVIA Lab with Prof. Dr. Mennatallah El-Assady
and PEACH Lab with Prof. Dr. April Yi Wang

I believe it is important to center, not undermine, human agency, creativity, and legacy in the age of AI.

  • Grind
    硏 /jʌn/ to polish
    stone grinded till even
  • Research
    究 구 /ku/ to research
    a group investigating a cave
AI and My Values: User Perceptions of LLMs' Ability to Extract, Embody, and Explain Human Values from Casual Conversations
*Anon*
20 people texted a chatbot for a month about their daily lives. The AI built profiles of their values, then explained its reasoning in a 2-hour interview. 13 participants left convinced the AI truly understood them... until we revealed the mechanics behind its 'empathy.'
AI and My Values: User Perceptions of LLMs' Ability to Extract, Embody, and Explain Human Values from Casual Conversations thumbnail 1AI and My Values: User Perceptions of LLMs' Ability to Extract, Embody, and Explain Human Values from Casual Conversations thumbnail 2AI and My Values: User Perceptions of LLMs' Ability to Extract, Embody, and Explain Human Values from Casual Conversations thumbnail 3AI and My Values: User Perceptions of LLMs' Ability to Extract, Embody, and Explain Human Values from Casual Conversations thumbnail 4AI and My Values: User Perceptions of LLMs' Ability to Extract, Embody, and Explain Human Values from Casual Conversations thumbnail 5AI and My Values: User Perceptions of LLMs' Ability to Extract, Embody, and Explain Human Values from Casual Conversations thumbnail 6AI and My Values: User Perceptions of LLMs' Ability to Extract, Embody, and Explain Human Values from Casual Conversations thumbnail 7
Does My Chatbot Have an Agenda? Understanding Human and AI Agency in Human-Human-like Chatbot Interaction
*Anon*
22 adults chatted with Day, our AI companion, for a month. Who decided when to greet, change topics, or say goodbye? Participants thought they were in control, but the AI was quietly steering depth and breadth. After revealing Day's strategies in interviews, we learned that agency isn't owned: it's negotiated, turn by turn.
Does My Chatbot Have an Agenda? Understanding Human and AI Agency in Human-Human-like Chatbot Interaction thumbnail 1Does My Chatbot Have an Agenda? Understanding Human and AI Agency in Human-Human-like Chatbot Interaction thumbnail 2Does My Chatbot Have an Agenda? Understanding Human and AI Agency in Human-Human-like Chatbot Interaction thumbnail 3Does My Chatbot Have an Agenda? Understanding Human and AI Agency in Human-Human-like Chatbot Interaction thumbnail 4Does My Chatbot Have an Agenda? Understanding Human and AI Agency in Human-Human-like Chatbot Interaction thumbnail 5
From Junior to Senior: Allocating Agency and Navigating Professional Growth in Agentic AI–Mediated Software Engineering
*Anon*
Juniors code with AI from day one. Seniors learned the hard way, then adapted. We studied both through debugging tasks and interviews, comparing how they delegate to agentic tools like Cursor. The gap isn't just skill, it's a completely different relationship with control, learning, and what it means to be a programmer.
From Junior to Senior: Allocating Agency and Navigating Professional Growth in Agentic AI–Mediated Software Engineering thumbnail 1From Junior to Senior: Allocating Agency and Navigating Professional Growth in Agentic AI–Mediated Software Engineering thumbnail 2
Generative AI in Knowledge Work: Design Implications for Data Navigation and Decision-Making
Published in CHI '25 · Honorable Mention Award (Top 5%)
Bhada Yun*, Dana Feng*, Ace Chen, Afshin Nikzad, Niloufar Salehi
Product managers drown in scattered information across platforms, so we built Yodeai to help them synthesize it all. 16 PMs tested it for real decisions. They praised the adaptability and control, but hit limits: overreliance, isolation, and blind spots AI couldn't see.
Generative AI in Knowledge Work: Design Implications for Data Navigation and Decision-Making thumbnail 1Generative AI in Knowledge Work: Design Implications for Data Navigation and Decision-Making thumbnail 2Generative AI in Knowledge Work: Design Implications for Data Navigation and Decision-Making thumbnail 3Generative AI in Knowledge Work: Design Implications for Data Navigation and Decision-Making thumbnail 4
Wrapped in Anansi's Web: Unweaving the Impacts of Generative-AI Personalization and VR Immersion in Oral Storytelling
Published in AH '25 (Augmented Humans)
Carrie Lau, Bhada Yun, Samuel Saruba, Efe Bozkir, Enkelejda Kasneci
We built a VR experience of Ghanaian Anansi folktales with AI-driven personalization to preserve oral traditions. 48 participants tried it. VR boosted their cultural interest as expected, but the AI personalization did something unexpected: it turned their focus inward, sparking self-reflection more than cultural learning.
Wrapped in Anansi's Web: Unweaving the Impacts of Generative-AI Personalization and VR Immersion in Oral Storytelling thumbnail 1Wrapped in Anansi's Web: Unweaving the Impacts of Generative-AI Personalization and VR Immersion in Oral Storytelling thumbnail 2Wrapped in Anansi's Web: Unweaving the Impacts of Generative-AI Personalization and VR Immersion in Oral Storytelling thumbnail 3

If you'd like to work with me, please message me on LinkedIn or email me at bhayun@ethz.ch.

© ∞ Bhada Yun