Assessing Individual Differences In Graphical Perception

Davis, Russell

Etd

Assessing Individual Differences In Graphical Perception

Public Deposited

Data is often presented in the form of graphical visualizations rather than as raw data, with encodings frequently chosen to optimize for accuracy of interpretation by the audience. Visualization guidelines have been drafted to help designers select visualizations that optimize the reader’s ability to understand it. However, most visualization guidelines are derived from studies that focus on population-level rankings of accuracy, disregarding possible individual differences in peoples ability to interpret visualizations. This thesis considers variations in individual performance by replicating and extending Cleveland & McGill’s widely-studied visualization experiment. By implementing Bayesian multilevel regression, we generate models that facilitate exploration of differences between individual participants and between each visualization type. We confirm that a substantial percent- age of individuals show accuracy judgments that deviate from the canonical rankings. We discuss between-individual differences as a relevant factor for design effectiveness, with respect to its capacity to highlight individual variation from population-level aggregates, and with respect to its ability to differentiate factors to between-individual variation; implications for research focused on providing guidance to visualization designers; and proposed further modifications to research in the mode of Cleveland & McGill.

Creator