Abstract

Recent advances in machine learning and deep learning have provided substantial gains in many predictive tasks like image recognition, text classification, and speech recognition. However, implementing such models in practice, especially for decision-making, can be difficult as these models are inherently opaque, non-intuitive, and difficult to interpret. This talk will explore the role of explainable systems built on top of such models to enable informed decision-making processes. Originally introduced by DARPA, the U.S. Department of Defense’s research arm, Explainable Artificial Intelligence (XAI) is an emerging field that aims to create machine learning techniques that (1) produce more explainable models while maintaining high predictive accuracy and (2) enable human users to understand, trust, and manage the outcomes from such models.