Learning Interpretability for Visualizations using Adapted Cox Models through a User Experiment

TL;DR

This paper investigates how to improve visualization interpretability by modeling user preferences through adapted Cox models, showing that neighborhood conservation measures outperform traditional cluster separability measures, especially when combined.

Contribution

It introduces a user-based approach to evaluate interpretability measures and demonstrates the effectiveness of combining multiple measures for better prediction of user preferences.

Findings

Neighborhood conservation measures outperform cluster separability measures.

Combining multiple interpretability measures enhances prediction accuracy.

User preferences are better modeled with neighborhood-based measures.

Abstract

In order to be useful, visualizations need to be interpretable. This paper uses a user-based approach to combine and assess quality measures in order to better model user preferences. Results show that cluster separability measures are outperformed by a neighborhood conservation measure, even though the former are usually considered as intuitively representative of user motives. Moreover, combining measures, as opposed to using a single measure, further improves prediction performances.