Impact of Explanations for Trustworthy and Transparent Artificial Intelligence

Abstract

Trust is a fundamental aspect in the interaction between humans and artificial intelligence (AI). Building and maintaining trust requires designing AI systems that are transparent, explainable and trustworthy, and providing appropriate feedback to users to ensure that they can understand the systems behaviour. This work aims at evaluating the impact of different explanations (local and global) on human’s trust and understanding of a facial expression recognizer. Results show that explanations are appreciated when present, but when no explanations are given, users apply their own mental model on how does the system work and trust it if their experience using it is positive.

Ähnliche Arbeiten