Computational Modeling for Ballot Design

2022-2024

Design choices in digital interfaces can have critical real-world consequences. In the 2018 U.S. Senate race in Broward County, Florida, nearly 31,000 voters failed to cast a vote, likely due to poor ballot design. This discrepancy may have directly altered the outcome of the election. Such examples highlight the urgent need to understand how visual design influences user behavior, especially in time-sensitive, high-pressure contexts like voting, healthcare, and emergency response.

This project investigates how users perceive and organize information on a screen, focusing on how Gestalt principles, such as proximity, similarity, and connectedness, affect visual grouping. Our findings emphasize that these cues do not act in isolation. Instead, they interact in complex ways depending on spatial layout, task demands, and user expectations. By isolating and testing these interactions, we aim to refine our understanding of how people navigate and interpret interfaces under pressure.

We have developed a computational model that predicts how users group interface elements. While grounded in ballot analysis, this model is extensible to any interface where clarity and efficiency are critical, from electronic health records to cockpit dashboards. Ultimately, this work contributes to a broader theory of human-centered design that reduces errors and improves decision-making through better interface structure.

This work was conducted under the guidance of Dr. Michael Byrne at the Computer-Human Interaction Lab with Shrreya Aagarwal. This work was presented at HFES ‘24 and AHFE ‘25.