Evidence for effort prediction in perceptual decisions
- Nisheeth Srivastava, Dept of Computer Science, Indian Institute of Technology, Kanpur, UP, India
AbstractThe classic drift diffusion model of the 2AFC choice process assumes that observers select evidence accumulation thresholds to optimize some desired level of accuracy across the experiment. We argue that it is more ecologically natural to assume that decision-makers set this threshold adaptively, using information from recent trials to adjust it for upcoming ones. To test this hypothesis, we designed and conducted a pair of random dot motion discrimination experiment where the coherence parameter that controls task difficulty varies across trials in a predictable manner. To analyze data from these experiments, we also designed a hierarchical drift diffusion model that allows decision-makers to adapt their evidence threshold based on the trend of difficulty of previous trials. Our results suggest that observers rationally integrate cross-trial information about trial difficulty into perceptual decision-making by adjusting their internal evidence thresholds. We briefly discuss the implications of the existence of such trial-level effort inference on contemporary models of the choice process.
Return to previous page