Purchase this article with an account.
Kevin Smith, Edward Vul; Physical prediction biases are faithful physics plus visual uncertainty. Journal of Vision 2013;13(9):777. doi: 10.1167/13.9.777.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To plan actions in the world, people must account for how it will change over time: for instance, catching a ball requires predicting its trajectory. People do this easily and flexibly, reflecting an underlying knowledge of physics, but they show task specific biases in applications of these physical principles. We propose that people use one physical model across tasks, but different sources of uncertainty interact with task demands to produce such task-specific biases. Each participant predicted where a pendulum bob would fall after the string had been cut in two qualitatively different tasks. In the ‘catching’ task, the computer chose when to cut the string, and we measured where participants moved a bucket horizontally to catch it. In the ‘cutting’ task, we measured when participants cut the string to make the bob fall into a bucket placed by the computer. All participants performed the same trials in each task, and showed high inter-subject reliability (split-half correlations – catching: r=0.993, cutting: r=0.998). Trials were matched across tasks such that where the ball landed in a ‘catching’ task was the bucket position in the matched ‘cutting’ trial. Thus, these tasks employ the same physical system, but are differentially sensitive to uncertainty in the location and speed of the bob, and elicited distinct biases: in the ‘catching’ task, participants placed the bucket closer to the center than the true landing position, while in the ‘cutting’ task, participants tended to overshoot the bucket. Aggregate participant responses in both tasks were well predicted by a physical model (catching: r=0.994, cutting: r=0.992) that differed across task only in which sources of visual uncertainty were relevant. These results suggest that people use a single model of the physical world, but that biases can occur due to the interactions of perceptual uncertainty and task demands.
Meeting abstract presented at VSS 2013
This PDF is available to Subscribers Only