Office of Undergraduate Research

Evaluating Hallucinations in AI-Generated Storyboards for Human-Centered Design DUIRI - Discovery Undergraduate Interdisciplinary Research Internship Summer 2026 Accepted Human–Computer Interaction; Artificial Intelligence; Design Research; Computing Education Text vignettes are carefully designed research instruments used to study people’s perceptions and decisions. When AI tools generate visual storyboards from these vignettes, they may introduce unintended details (“hallucinations”) that change the meaning of the original scenario. This project investigates how faithfully AI-generated storyboards preserve the controlled elements of vignette-based studies. Students will help compare original text scenarios with AI-generated visuals, identify where meaning shifts occur, and evaluate how these changes could bias research findings. The goal is to develop guidelines for using AI responsibly in human-centered research and design. Tianyi Li Shu Hu Students will contribute to:
Generating AI storyboards from carefully controlled research vignettes
Systematically comparing visual outputs with original experimental scenarios
Coding deviations that alter meaning, context, or participant interpretation
Evaluating how these deviations may threaten validity of research instruments
Assisting with analysis, visualization, and research documentation
Students will gain experience in HCI research methods, trustworthy AI evaluation, and experimental design. GPA ? 3.0 Sophomore or above preferred Background in Computer Science, HCI, Design, Data Science, or related fields Ability to read research papers and implement methods described in them Strong programming and analytical skills (e.g., Python or similar) Comfortable working in a fast-paced research environment with iterative deadlines Strong attention to detail and ability to follow carefully controlled research protocols 0 20 (estimated)