Bandits for Learning to Explain from Explanations
This work addresses the problem of jointly learning predictions and explanations for system designers, offering an incremental approach.
This paper introduces Explearn, an online algorithm that learns to jointly provide predictions and their corresponding explanations. It utilizes Gaussian Process-based contextual bandits to capture different explanation types and control generalization, with initial experiments showing promise.
We introduce Explearn, an online algorithm that learns to jointly output predictions and explanations for those predictions. Explearn leverages Gaussian Processes (GP)-based contextual bandits. This brings two key benefits. First, GPs naturally capture different kinds of explanations and enable the system designer to control how explanations generalize across the space by virtue of choosing a suitable kernel. Second, Explearn builds on recent results in contextual bandits which guarantee convergence with high probability. Our initial experiments hint at the promise of the approach.