Visualizations are widely used to compare aggregate statistics between subsets of data. However, aggregation can often obscure patterns or trends and produce misleading views of the data. One example of this risk is Simpson’s Paradox, a phenomenon that can commonly occur in interactive data visualizations that enable ad hoc grouping and filtering. We explore the potential of counterfactuals— widely used in causal inference— to help counter the risks of invalid conclusions due to Simpson’s Paradox in data visualization.