Counterfactual Fairness Does Not Imply Group Fairness in Image Classification
July 9, 2026
A study using newly constructed datasets, OursCeleb and OursLFW, demonstrates that counterfactual fairness does not guarantee group fairness in image tasks. This contradicts previous findings in tabular datasets and highlights a disconnect in how fairness metrics interact in visual domains.
HOW THIS AFFECTS YOU
●
researcherYou must account for the divergence between counterfactual and group fairness when evaluating vision models.
●
policyThis suggests that current fairness benchmarks may provide a false sense of security regarding algorithmic bias.