Tim Miller

Tim connects AI quality with people's ability to inspect and challenge a system's reasoning. Visual Evaluative AI gives users assessing skin images a way to examine evidence for and against a possible diagnosis. Collaborative research on AI-generated study participants also identifies both potential and variability in explanation testing, offering useful evidence for researchers evaluating decision-support software.

Tim Miller is TIET-UQ Chair of Data Science and Professor of Artificial Intelligence at the University of Queensland, leading the Augmented Socio-Cognitive Intelligence lab. The lab combines machine learning, interaction design and cognitive science to help people make better decisions. Previously a Professor of Computer Science at the University of Melbourne and founding co-director of its Centre for AI and Digital Ethics, Tim remains an honorary professor there.

Much of this scholarship focuses on explainable AI to help people understand and assess automated outputs. With collaborators, Tim developed Visual Evaluative AI, a decision aid that presents evidence for and against a proposed explanation of an image. Evaluated in the skin cancer domain, the web-based prototype lets users upload a skin image, select a possible diagnosis and examine evidence expressed through human-understandable concepts.

Collaborative studies also examine how explanations are tested, including an experiment using seven large language models to reproduce human responses in an explanation evaluation. Most original conclusions were replicated, though agreement varied by model and experimental settings. Together, these projects connect software quality with people's ability to scrutinise AI-supported decisions and with rigorous evaluation of the explanations provided.