Bayesian Modelling of Alluvial Diagram Complexity



Anjana Arunkumar, Shashank Ginjpalli, Chris Bryan

 External link (DOI) 

 View presentation:2021-10-28T15:20:00ZGMT-0600Change your timezone on the schedule page
2021-10-28T15:20:00Z

Exemplar figure, described by caption below — We study the complexity of alluvial diagrams with the following workflow: (1) We generate a publicly-available testing dataset of alluvial diagrams that vary complexity based on a combination of features. (2) We conduct two crowdsourced user studies and (3) assess the task performance and perceived complexity of the charts. (4) We then model the complexity of alluvial diagrams with a combination of regression and factor analysis and by constructing Bayesian models.

Fast forward

Direct link to video on YouTube: https://youtu.be/ugnuGYwf3gE

Keywords

Perception & Cognition, Datasets, Human-Subjects Quantitative Studies

Abstract

Alluvial diagrams are a popular technique for visualizing flow and relational data. However, successfully reading and interpreting the data shown in an alluvial diagram is likely influenced by factors such as data volume, complexity, and chart layout. To understand how alluvial diagram consumption is impacted by its visual features, we conduct two crowdsourced user studies with a set of alluvial diagrams of varying complexity, and examine (i) participant performance on analysis tasks, and (ii) the perceived complexity of the charts. Using the study results, we employ Bayesian modelling to predict participant classification of diagram complexity. We find that, while multiple visual features are important in contributing to alluvial diagram complexity, interestingly the most important feature changes depending on the type of complexity being modeled, i.e. task complexity vs. perceived complexity.