Current Artificial Intelligence (AI) and Machine-Learning (ML) based systems lack transparency. Transparency is a design principle that is consistently presented in the literature as a means to improve human performance with AI and ML based systems. This talk will present findings from an experiment exploring the impacts of automation transparency on human decision making in ML-based decision support systems and a meta-analysis of automation transparency. These empirical and statistical examinations demonstrate little evidence that automation transparency is a generalizable principle. Even in similar contexts, variations in implementing this principle have led to contradictory results. These uncertainties have safety and efficiency implications for those implementing transparency interventions for autonomous and Machine-Learning based systems.
Dr. Fahimeh Rajabiyazdi is a Human Factors Designer at Candu Energy – a member of AtkinsRéalis. She has 7+ years of experience as a user experience and human factors researcher. She received a Doctor of Philosophy in Human Factors Engineering from the University of Toronto and a Master of Science in Human-Computer Interaction and Design from the KTH Royal Institute of Technology and Université Paris-Sud. She is passionate about designing automated systems and Machine Learning-based systems that improve human decision-making performance.