Adaptive group integration rules in a signal detection task
36.44, Sunday, May 12, 2:45 - 6:45 pm, Orchid Ballroom
Mordechai Z. Juni1, Miguel P. Eckstein1; 1University of California, Santa Barbara
Previous studies on collective decision-making have investigated what rules of integration people adopt, and whether the groups collective performance outperforms each members individual performance. Here we explore whether groups could beneficially adapt their perceptual integration rule when necessary. Triads participating simultaneously on separate computers were challenged to detect a Gaussian signal (SD=0.5 DVA) embedded in white noise (signal present 50% of the time), both individually and collectively. On each of 400 trials, participants first recorded and announced their personal ratings as to whether the signal was absent (1, 2, 3, 4, 5) or present (6, 7, 8, 9, 10). Next, they freely discussed what their collective rating should be (given their individual ratings) until they agreed upon a group rating. Finally, feedback was provided as to whether the signal was present on that trial. During the first 200 trials, the signal (when present) had 2% contrast on all three computers. During the last 200 trials, the signal (when present) had 0.5% contrast on two computers (randomized on each trial) and 9% contrast on the remaining computer. While all participants (nine in total) noticed that the signal was occasionally quite strong during the second half of the experiment, none of them reported realizing (upon subsequent questioning) that the three members of the group were being shown signals of different strengths during the second half. Nevertheless, while initially the group ratings were more correlated with an average rating rule (p < .05 for all three groups), they slowly adapted during the second half until finally, during the last 100 trials, the group ratings were more correlated with a more optimal rule approximated by a follow the most extreme rating rule (p < .05 for all three groups). This indicates that groups flexibly adopt different perceptual integration rules to improve collective performance.