This academic article presents a new approach to counterfactual reasoning in fairness and recourse within artificial intelligence (AI) systems. Traditionally, counterfactual reasoning involves imagining hypothetical interventions on legally-protected characteristics such as race or gender. This method, however, has limitations. The authors propose a new paradigm based on backtracking counterfactuals. Instead of imagining interventions, this approach considers alternate initial conditions while holding the protected characteristics constant. This allows for the consideration of the same social concerns but without relying on demographic interventions.
Publication date: 26 Jan 2024
Project Page: Unavailable
Paper: https://arxiv.org/pdf/2401.13935
