Three Levels of AI Transparency

Abstract:
Transparency is generally cited as a key consideration towards building Trustworthy AI. However, the concept of transparency is fragmented in AI research, often limited to transparency of the algorithm alone. While considerable attempts have been made to expand the scope beyond the algorithm, there has yet to be a holistic approach that includes not only the AI system, but also the user, and society at large. We propose that AI transparency operates on three levels, (1) Algorithmic Transparency, (2) Interaction Transparency, and (3) Social Transparency, all of which need to be considered to build trust in AI. We expand upon these levels using current research directions, and identify research gaps resulting from the conceptual fragmentation of AI transparency highlighted within the context of the three levels.