Bayesian-Assisted Inference from Visualized Data

A Bayesian view of data interpretation suggests that a visualization user should update their existing beliefs about parameter's value in accordance with the amount information parameter captured by new observations. Extending recent work applying models to understand and evaluate belief updating from visualizations, we show how predictions inference can be used guide more rational updating. We design inference-assisted uncertainty analogy numerically relates observed user's subjective uncertainty, posterior prescribes given prior data. In pre-registered experiment on 4,800 people, find when newly sample is relatively small (N=158), both techniques reliably improve people's average compared current best practice visualizing For large samples (N=5208), where updated tend deviate strongly prescriptions model, evidence effectiveness two forms assistance may depend proclivity toward trusting source discuss our results provide insight into individual processes understanding these aspects paves way for sophisticated interactive visualizations analysis communication.