Enabling Effective Transparency: Towards User-Centric Intelligent Systems

Abstract

Much of the current research in transparency and explainability is highly technical and focuses on how to derive explanations from models and algorithms. Less thought is being given to how users actually want to receive transparency and explanations from intelligent systems. My work tackles transparency and explainability from a user-centric perspective. I examine why transparency is desirable by showing that users may be susceptible to deception from intelligent systems. I demonstrate when users want transparency. Finally, my work begins to uncover how users want transparency conveyed. This body of work intends to create a path for designing transparency that puts user needs first rather than creating transparency as a convenient afterthought of model selection.