Abstract
Predicting the future state of natural and social events is crucial for our adaptive behaviors. Whereas the sensory system predicts incoming inputs over a range of several hundred milliseconds, humans are governed by a variety of cognitive biases in predicting the future beyond a few seconds. To elucidate such biases, we investigated the characteristics of human prospective decision making upon the supra-second future of a random-walk stimulus. In our psychophysical procedure, a single Gabor patch (1.0 c/deg) changed horizontal position in accordance with a 1/f velocity noise (with cut-off below 4.7 Hz) over variable periods of time (T = 1, 2, or 4 sec). Average velocity over the presentation was set zero (i.e., no linear trend). After the stimulus disappeared, observers indicated within 0.5 sec whether the stimulus would move toward left or right over a particular period after stimulus offset (δT= 1, 2, 4 sec for each block). We applied logistic regression analysis to calculate the impact of velocity information at each temporal frame upon the observer's response. The analysis revealed that observer decisions tend to depend heavily on velocities at and soon before stimulus offset, especially after the last several reversals in motion direction. This tendency becomes less marked when the observers tried to predict events in the relatively far future (δT = 4 sec). These results support the notion that, when predicting the uncertain future, humans tend to focus heavily on the last few directional reversals just before they make a decision and to linearly extrapolate trajectory to estimate the future state.
Meeting abstract presented at VSS 2018