AJ Jansen¹, Brett Fajen¹; ¹Rensselaer Polytechnic Institute

Action capabilities pose critical constraints on how people move through the world. While running, biking, or driving, humans can only accelerate so quickly and turn so sharply. Previous work shows that people can accurately perceive whether actions are within their capabilities and adjust their behavior during continuous control tasks to accommodate their limits. These findings raise questions about how people initially learn and continually relearn their ever-changing action capabilities. When action capabilities are enhanced such that new, potentially more efficient solutions become possible, do people spontaneously switch to those solutions or stick with well-practiced solutions? To address this question, we conducted two experiments using a multiple-waypoint steering task performed in a virtual environment. Subjects guided a drone through two gates at a fixed distance on the centerline and a third gate at a variable distance and angle. Each block comprised 18 trials. The drone was sluggish in Blocks 1-5, gradually transitioned to agile in Block 6, and remained agile in Blocks 7-11. With a sluggish drone, the task could be performed most successfully by following a curved trajectory to the second gate that anticipated the third gate. With an agile drone, a more direct path was both possible and more time efficient. Nevertheless, no subjects immediately switched to the direct-path solution when it first became possible (Block 7) and only some subjects converged toward a direct-path solution by Block 11. In Experiment 2, we instead used an obstacle-avoidance task for Block 6 to test whether performing a different task during adaptation would promote switching solutions. As in Experiment 1, subjects tended to stick with the indirect-path solution that they learned when the drone was sluggish. These findings suggest that when action capabilities change, people may (at least initially) stick with well-practiced solutions even if more time-efficient solutions are available.

Acknowledgements: NSF 2218220