Counterfactuals and Visual Causal Inference

Research projects within this theme examine and evaluate new visual analytics methods that enhance the accuracy and robustness of causal inferences drawn by users of data visualizations. Our research (along with studies from other labs) has shown that people often draw causal inferences from patterns displayed in visualizations even when such inferences are unfounded. We aim to develop new methods which align more closely with how people use visualizations to interpret data so as to help support improved understanding of which patterns are likely to indicate meaningful causal relationships. Our research also studies how people understand causality based on both traditional visualizations and our own new methods.

Our research exploring counterfactual-based supports for visual causal inference is supported in part by National Science Foundation Award #2211845.


Software Products

Certain methods developed through this research initiative are embodied within the Co-Op software library developed by our lab. This library is available in both Python and JavaScript versions for use in visualizations developed on both platforms. Our research exploring counterfactuals and causal inference also inspired the development of the CausalSynth web application for synthetic data generation from ground-truth causal models. More information and links to both software products’ open-source GitHub repositories can be found on our lab’s software page.