Poster F130, Tuesday, March 28, 8:00 – 10:00 am, Pacific Concourse
Decoding the Representational Space of Decision Values using EEG
Pablo Morales1, Atsushi Kikumoto1, Ulrich Mayr1; 1University of Oregon
Losses loom larger than gains during decision making––likely leading to decision phenomena such as risk aversion or the status-quo bias. We do not have a full understanding about how gain and loss information is represented and how these representations contribute to decision making. In order to probe the representational space of value information we recorded EEG during a novel gambling task: Trials consisted of streams of eight stimuli, each indicating monetary values along a continuous dimension ranging from $0-$20. Subjects were asked to choose between a random pick from one of these eight stimuli or a sure gain of $10. Behavioral choices indicated typical risk aversion tendencies. Decoding of time-frequency decomposed EEG via a linear classifier revealed that oscillations in the delta range (~1-3Hz) carried information that reliably indexed each stimuli’s gain/loss qualities (relative to the sure gain option). A median split of the sample into individuals with high versus lower risk aversion revealed a sharp, categorical gain-loss representation in risk-averse subjects. Less risk-averse subjects showed a graded pattern indicative of a more continuous gain-loss representation. These results indicate that risky choices are driven in part by the precision with which value information is represented, where course gain/loss distinctions lead to stronger risk aversion. In contrast, there was no evidence for a greater sensitivity to losses on the level of the observed value representations (i.e., a shift of the category boundary), suggesting that the actual risk aversion is created upstream from these representations.
Topic Area: THINKING: Decision making