Machine Learning and Knowledge Extraction

Explaining sophisticated machine-learning based systems is an important issue at the foundations of AI. Recent efforts have shown various methods for providing explanations. These approaches can be broadly divided into two schools: those that provide a local and human interpreatable approximation machine learning algorithm, logical exactly characterise one aspect decision. In this paper we focus upon second school exact explanations with rigorous foundation. There epistemological problem these methods. While they furnish complete explanations, such may too complex humans to understand or even write down in readable form. Interpretability requires epistemically accessible grasp. Yet what sufficiently explanation still needs clarification. We do here terms counterfactuals, following [Wachter et al., 2017]. With counterfactual many assumptions needed are left implicit. To so, exploit properties particular data point sample, as also well partial explore how move from call then global ones. But preserve accessibility argue need partiality. This partiality makes it possible hide explicit biases present algorithm injurious unfair.We investigate easy uncover fair by exploiting structure set counterfactuals explanation.