Abstract
<p>Risky options often differ not only in expected value and risk but also in complexity. Although complexity aversion is well documented in the gain domain, it remains unclear whether this aversion extends to losses. We contrast two theoretical accounts that make diverging predictions: if complex options are avoided because processing them requires cognitive effort, complexity aversion should be similar for gains and losses; if complexity instead induces cognitive uncertainty, whose evaluation depends on domain-specific risk preferences, complexity aversion should be weaker for losses than for gains. We tested these predictions in two preregistered experiments (N = 138 and 135) using gambles that vary outcome complexity, separately for the gain and loss domain: Study 1 examined choices between two risky options, and Study 2 contrasted a risky with a sure option. In both studies, participants chose complex options significantly less often in the gain than in the loss domain; specifically, they were complexity averse for gains but complexity neutral for losses. Response times showed a converging asymmetry, significant in Study 2. Subjective ratings of cognitive uncertainty predicted complexity aversion in the gain but not in the loss domain. These results favor the cognitive-uncertainty account over the effort account: complexity increases cognitive uncertainty, whose behavioral consequences depend on the outcome domain.</p>