Fair Conformal Classification via Learning Representation-Based Groups
For machine learning practitioners concerned with fairness, this method provides a practical way to ensure equalized coverage across subgroups in conformal prediction, addressing algorithmic bias.
This paper introduces a fair conformal inference framework for classification that constructs prediction sets with conditional coverage guarantees on adaptively identified subgroups, balancing compactness and equalized coverage. Experiments on synthetic and real-world datasets demonstrate the framework's effectiveness.
Conformal prediction methods provide statistically rigorous marginal coverage guarantees for machine learning models, but such guarantees fail to account for algorithmic biases, thereby undermining fairness and trust. This paper introduces a fair conformal inference framework for classification tasks. The proposed method constructs prediction sets that guarantee conditional coverage on adaptively identified subgroups, which can be implicitly defined through nonlinear feature combinations. By balancing effectiveness and efficiency in producing compact, informative prediction sets and ensuring adaptive equalized coverage across unfairly treated subgroups, our approach paves a practical pathway toward trustworthy machine learning. Extensive experiments on both synthetic and real-world datasets demonstrate the effectiveness of the framework.