Description
Researchers are conducting a study to see if experts in important jobs, like doctors and military personnel, can make better decisions alongside data-driven predictions from artificial intelligence (AI). The study focuses on giving these experts more detailed information about the AI prediction, such as how confident the AI is in its predictions, to see if it helps them work better with the AI system. The study involves two groups of experts: radiologists who examine X-rays for possible cancers and members of the Royal Canadian Navy (and other allied nations) who look for underwater mines using sonar images. The goal of the study is to determine how to best show important AI information to expert users who make safety-critical decisions. The researchers expect that showing different factors of the AI confidence information using principles from Ecological Interface Design, specifically in relation to the representativeness of the images within the AI's training datasets, will help these experts make better decisions when using AI for their tasks. The researchers hope that by doing this, they can figure out the best way to show AI predictions to help professionals in safety-critical jobs know when to trust the AI system and when to potentially ignore the system and rely on their expertise. Ultimately, this could lead to safer decisions in these important fields if (or when) AI is introduced, as well as identify opportunities for leveraging research knowledge between two safety-critical professions in the future design of human-centered AI systems.
Research Partners
Defence Research and Development Canada - Atlantic Research Centre