This is a past event.

The next ACSHF Forum will be on Monday, January 27 in room 109 of the Harold Meese Center.  There will be two speakers: Lamia Alam and Anne Linja, both ACSHF Graduate students.

Lamia Alam will be presenting Physicians' Explanation Strategies in Re-diagnosis Scenarios. Explanation in Artificial Intelligence (AI) requires an understanding of human explanatory reasoning.  An effective way to comprehend both AI and human explanations for decision making is to demonstrate them in the medical diagnosis domain. The uncertain nature of clinical decisions often leads physicians to go through re-diagnosis and they need to communicate and explain these to their patients. This presentation will discuss a recent study exploring physicians’ explanation strategies in these scenarios by conducting ACTA (Applied Cognitive Task Analysis)-based interviews with physicians. Five broad themes: 1) Initial steps, 2) Knowing the audience, 3) Using case information, 4) Tools to support data/fact, and 5) Emotional Aspects were identified from these interviews. A generic meta-timeline of fitting explanations in re-diagnosis cases is also design based on the observation from the interviews. The study also indicates what areas of explanation AI should concentrate on to improve user trust and satisfaction for medical diagnosis. 

Anne Linja will present Understanding Driver Behavior: Filling in the Human Factors Gap for the Short Storage Gap. Train-vehicle collisions at highway-rail grade crossings continue to be a safety concern with more than 2000 rail crossing incidents each year in the United States alone; twenty percent of these occur at short-storage rail crossings. Many of these incidents are attributed to errors in human judgment. Despite improvements in warnings, and driver's awareness of safe driving at rail crossing, the frequency of incidents has not decreased. In this exploratory study, we examined safety concerns at rail crossings. In experiment 1, twenty-nine experienced drivers (50% from rural settings) viewed 10 driving
scenes and identified, rated, and explained up to five safety concerns in each image. Overall 1477 safety concerns were identified, coded, and analyzed. In experiment 2, twenty participants were eye-tracked while analyzing the same images for safety concerns. Initial results indicate participants were sensitive to the safety concerns. This MS work extends prior work through a new experimental paradigm and is part of a larger study examining strategies for improving rail safety. This project is being done in collaboration with the MTU Rail Transportation Program