Decision-making under risk is often studied with fully described lotteries, where normative theory predicts that post-choice outcome disclosure (feedback) should not influence preferences. However, previous empirical work has generally shown that feedback does affect risk-taking, yet, without reaching a consensus on the consequences of feedback or the underlying cognitive mechanisms. Here, across seven behavioral experiments, we disentangle two competing accounts: the learning hypothesis, where feedback alters subjective values through experience, and the attitudinal hypothesis, where feedback changes preferences in anticipation of outcomes. We find that feedback does not improve maximization but consistently increases risk-taking. Fine-grained temporal analyses reveal that this effect emerges before any outcomes are experienced, ruling out learning as the primary driver. Moreover, the increase of risk-taking in partial feedback seems to be driven by curiosity, while in complete feedback by anticipated regret. Our results indicate that feedback can bias decision-making primarily through attitudinal rather than learning mechanisms.