Computer Science Department , PhD Proposal Defense Akim Ndlovu "Advancing the Design and Evaluation of Thematic Map Visualization Techniques "

Akim Ndlovu 

PhD Candidate 

WPI – Computer Science Department 

Monday, May 19, 2025 

Time: 1:00 p.m. – 2:00 p.m. 

Location : Fuller Labs 141 

Committee members:

Prof. Lane Harrison, Advisor,  WPI – Computer Science Department 

Prof. Charlie Roberts,WPI – Computer Science Department 

Prof. Emmanuel Agu, WPI – Computer Science Department 

Dr. Evan Peck, Associate Professor, UC Boulder (External Advisor)

Abstract: 

Thematic maps are commonly used for visualizing spatial data. Epidemiologists use choropleth maps to monitor and disseminate information related to disease outbreaks, journalists use bivariate choropleth maps to visualize the relationship between voter share and medicaid enrollment rates, and federal analysts use surprise maps to highlight regions with unexpected population changes. However, well-known challenges remain in the design and use of thematic maps: 1) choropleth maps are associated with known biases (i.e. the effects of area, color, and population in the interpretation of event rates for map exploration tasks), 2) model-derived maps—such as surprise maps—can require extensive parameter setting and tuning for effective use, and 3) visualizing multiple variables using geospatial maps and interpreting their spatial relationships remains significantly challenging.

 In this work, we address these challenges by 1) conducting empirical studies to evaluate how people interpret different mapping techniques, and 2) developing a system (i.e. SurpriseExplora) that enables users to make discoveries in spatial data that are

difficult to identify with choropleth or static surprise maps alone. Our findings offer a clearer understanding of how different map representation techniques influence the metrics people consider when exploring maps and how well-known biases shape the interpretation of the presented metrics. Expert reviews highlight the utility of SurpriseExplora and yield opportunities for future visualization toolkit design. In the proposed work, we aim to improve how people use geospatial maps to interpret multivariate spatial data. We propose adopting a new think-aloud crowd-sourced technique to evaluate how people make sense of either univariate or bivariate maps. We anticipate these studies will reveal gaps in our understanding of how people interpret the increasingly popular bivariate colorscale technique and lead to recommendations for the design and use of bivariate choropleth maps in practice.