| Abstract No.: | C-G3193 |
| Country: | Canada |
| | |
| Title: | MAKING DECISIONS AS THE EVIDENCE IS CHANGING |
| | |
| Authors/Affiliations: | 1 Stephany El-Murr; 1 Stephany El-Murr*; 1 Geneviève-Aude Puskas; 1 Geneviève-Aude Puskas; 1 Paul Cisek; 1 Paul Cisek;
1 Department of Physiology, University of Montreal, QC, Canada.
|
| | |
| Content: | Human behavior in a variety of decision-making tasks can be explained in terms of a class of theories called “sequential sampling” or “accumulator models”. These models suggest that the process of decision-making involves the accumulation of evidence for given options. A choice is then selected when the accumulated evidence for that choice reaches some fixed threshold. These models have been shown to successfully predict the distribution of reaction times in a variety of tasks in which a decision was made on the basis of evidence that remained constant in time. Here, we examine these models under conditions in which the evidence for or against given options is constantly changing. We presented subjects with a two-choice reach-decision task in which the evidence in favor of the choices gradually changed over time. Each trial consisted of 15 tokens moving one-by-one, every 200 ms, from a central pool to either a left or right target circle. The subject’s task was to select the target which they thought would receive the most tokens. Subjects were allowed to make their choice as soon as they felt confident, and at that point all the remaining tokens were quickly allocated and the decision evaluated. In “easy” trials, the evidence quickly favored one target over the other. In “ambiguous” trials, it remained ambiguous until later in the trial. In “misleading” trials, the evidence initially favored one target but then switched to favor the other. We also created “the bias against” trials where the first three tokens move in the incorrect target followed by the rest of the tokens moving in the correct target. In contrast, the “bias-for” are a mirror image of the “bias-against” trials for the six first tokens then are the same for the rest of the trial. We examined how differences in the temporal profile of the evidence influenced the subjects’ behavior. In particular, at each moment in time, we can compute the “instantaneous probability” of each choice being correct, which is determined by the distribution of tokens at that moment. A simple prediction is that decisions are made when instantaneous probability exceeds some threshold. However, accumulator models do not make the same prediction. In contrast, they predict that what determines the decision is not simply the probability at a given moment, but rather, the cumulative probability (i.e. the integral of probability). Thus, they predict that in longer, ambiguous trials, subjects will make their decision at a lower level of instantaneous probability than in the easier trials. Our results confirm this prediction, providing evidence that accumulator models with a certain degree of “leakage”, or loss of information can be used to explain behavior even during tasks in which stimulus evidence is constantly changing. |
| | |
| Back |
|
