Biography
Xinxuan Wang is an MSc User Experience Engineering student at Goldsmiths, University of London. Her academic interests focus on human-AI interaction, user trust, AI transparency, and the role of emerging technologies in design practice. With a background in design and UX research, she is particularly interested in how AI systems can support designers while also introducing new risks, such as inaccurate, overconfident, or misleading outputs.
Xinxuan Wang is an MSc User Experience Engineering student at Goldsmiths, University of London. Her academic interests focus on human-AI interaction, user trust, AI transparency, and the role of emerging technologies in design practice. With a background in design and UX research, she is particularly interested in how AI systems can support designers while also introducing new risks, such as inaccurate, overconfident, or misleading outputs.
Internship Project Summary
My internship project investigates AI hallucination and transparency in AI-assisted design tasks. The study focuses on how different levels of hallucination and epistemic transparency in AI-generated design suggestions may affect users’ trust, verification behaviour, and willingness to adopt the suggestions. The project uses a design scenario involving a smart medication reminder box for elderly users, where inaccurate or overconfident AI recommendations may create potential safety risks. The aim is to explore how clearer communication of uncertainty, assumptions, limitations, and verification needs can support more calibrated trust in AI-assisted design.
My internship project investigates AI hallucination and transparency in AI-assisted design tasks. The study focuses on how different levels of hallucination and epistemic transparency in AI-generated design suggestions may affect users’ trust, verification behaviour, and willingness to adopt the suggestions. The project uses a design scenario involving a smart medication reminder box for elderly users, where inaccurate or overconfident AI recommendations may create potential safety risks. The aim is to explore how clearer communication of uncertainty, assumptions, limitations, and verification needs can support more calibrated trust in AI-assisted design.