Tuesday, November 20, 2012 - 18:00 - 19:30
Mary Hayhoe (University of Texas)
While moving through natural environments, humans must accomplish a variety of behavioral goals, such as control of heading, obstacle avoidance, etc, which are based on uncertain knowledge about the state of the world and the body's relation to it. To accomplish these goals, humans must select the necessary visual information from the environment at the appropriate time, through control of gaze. The timing and choice of gaze targets and the accompanying attentional shifts are intimately linked with ongoing behavior. While there have been attempts to explain how gaze might respond to properties of the visual stimulus, we do not have a good understanding of how gaze targets are chosen for the purpose of achieving behavioral goals. In the context of a virtual driving task, we examine evidence for the hypothesis that gaze is prioritized in terms of a utility-weighted uncertainty metric, whereby fixations on task-relevant objects depend on both reward structure and task-relevant uncertainties.