Purchase this article with an account.
Brian Sullivan, Leif Johnson, Constantin Rothkopf, Dana Ballard, Mary Hayhoe; The effect of uncertainty and reward on fixation behavior in a driving task. Journal of Vision 2012;12(9):1259. doi: https://doi.org/10.1167/12.9.1259.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Most current models of human fixation behavior do not address the complex issue of how fixations are selected in dynamic, interactive natural tasks. Sprague et al (2007) proposed a model that explains task-based fixations as the outcome of arbitration between competing visual tasks. Fixation target selection uses a metric that weights reward estimates for task modules with uncertainty estimates for task relevant visual features. While reward has a known role in gaze allocation, there is little research addressing the role of state uncertainty, particularly in natural tasks. To address this, we manipulated rewards and task uncertainties while subjects drove in a virtual environment. Subjects drove through a virtual city in four conditions, which varied task importance and the presence of uncertainty. Implicit reward was manipulated by instructions that either emphasized driving an exact speed or following another car. Uncertainty was introduced by adding small random variations to the car’s speed and was reflected by the car’s speedometer. When exact speed was most important, adding uncertainty increased fixation proportion on the speedometer by 6%, and showed a trend to increase fixation durations by 80 ms and reduce inter-fixation intervals by 1s, over the same condition without uncertainty. This pattern was modulated in the following conditions where speedometer fixations were less important. When exact speed was most important with added uncertainty, the proportion of fixations on the speedometer increased by 16%, fixation duration increased by 170ms, and inter-fixation interval reduced by 6s over the follow condition with added uncertainty. These result suggest that when the state of a task-relevant visual variable becomes uncertain, subjects prioritize its gaze allocation, but only when this is a task with high reward. Thus both reward and state uncertainty are important components of gaze control in natural behavior, consistent with predictions from Sprague et al’s model.
Meeting abstract presented at VSS 2012
This PDF is available to Subscribers Only