Abstract
Uncertainty aversion is a preference for gambles with known risks over gambles with unknown risks (uncertainty). In a typical Ellsberg paradox experiment (Ellsberg, 1961), people prefer to bet on an urn with 50 green and 50 red marbles than on one with 100 marbles in which it is unknown how many of them are green and how many are red. We created two sequential search tasks, one abstract and one visual, whose outcomes are exactly the same as a lottery with, for example, a 70% probability of winning $10 and a 30% of winning nothing. In the abstract search, subjects saw a 10 × 10 grid of boxes and knew that 7 of them contained a green chip. They could open 10 boxes and finding at least one green chip would make them win $10. In the visual search, subjects had to look for Ts among Ls on a 1/f noisy background and were awarded $10 for correct answers. Half of the screens would contain one T, but the probability was unknown to subjects. By varying the opacity of the objects, it was possible to manipulate visual uncertainty and match, for every subject, the probability of a correct answer with the 70% probability of finding a green chip. Therefore, the abstract search was equivalent to a lottery with known probabilities (risk) and the visual search to one with unknown probabilities (uncertainty). While our subjects exhibited uncertainty aversion in the Ellsberg's paradigm, they did not prefer the abstract risky lottery to the visual uncertain lottery. Perceptual tasks have already been proven to produce nearly-optimal performance, as opposed to abstract lotteries in which people are consistently suboptimal (Maloney et al., 2006; Navalpakkam et al., 2009; Pedersini et al., 2010). Our results suggest that visual lotteries also contribute to eliminate cognitive biases like uncertainty aversion.
Office of Naval Research (MURI Program) to D.B.