Ziwei Nina Chen

My name is Ziwei Nina Chen. I am a first-year PhD student in Computer Science & Engineering at UC San Diego, advised by Prof. Kristen Vaccaro. I am also a member of the Design Lab. My current research focus is leveraging AI power to facilitate communication across roles (i.e. designers and engineers), and understanding potential harm caused by the use of LLMs (i.e. dark patterns). I am interested in researching, designing and building people-centered systems.

Before my PhD program, I worked as a UX designer at Dusty Robotics in Mountain View, where I designed robot control app for construction professionals to print digital layouts on the ground. I obtained my master’s degree in Information from the University of Michigan. At the UM, my research, advised by Prof. Florian Schaub, focused on addressing water access challenges faced by the homeless population. This project won first prize in the CHI 2024 Student Design Competition.

Email  /  CV  /  Scholar  /  Github

Recent work has shown that front-end code generated by large language models (LLMs) can embed deceptive designs. This study analyzes a large set of LLM-generated web components to identify deceptive designs and examines LLM design rationales through three prompts to understand their design decisions. We evaluate 15 commonly seen ecommerce components and generate a total of 1,080 components using four LLMs. We find that 55.8% of the generated components contain at least one deceptive design, with Interface Interference emerging as the dominant strategy, which uses color psychology to influence actions and hides essential information. The occurrence of deceptive designs varies across models, with DeepSeek producing the fewest. We find that prompts emphasizing company interests (e.g., increasing sales) significantly increase deceptive designs, whereas prompts incorporating human values effectively reduce them. Our findings highlight risks in using LLMs for coding and offer recommendations actions for LLM providers.

Access to clean water is essential, yet it poses a significant challenge for the homeless population. Our project, ’Where’s the Water,’ is a web-based tool designed to improve water access for the homeless community. It maps nearby clean water sources like drinking fountains, public restrooms, and showers. The tool’s design was informed through interviews with the homeless community in Ann Arbor, Michigan. The insights gained from these interviews were further supported by key findings from recent studies related to homelessness and water access. Besides locating, our tool’s functionality also includes filtering sources for operational hours and water quality. It features crowd-sourcing, allowing users to add new sources on the map, effectively utilizing community knowledge. In this article, we describe our research and design approach, highlighting the community and organizational feedback that helped turn our concept into a useful tool.

Template adopted from Jon Barron