Preference reversal: Memory and contextual biases with choice

A virtue of good theory is that it is general; theories that predict a wide range of events have obvious merit. Several prominent theories of decision making achieve this objective by proposing that all decisions can be modeled with the same generic representation (e.g.,
Birnbaum, 2008; Tversky & Kahneman, 1992; von Neumann & Morgenstern, 1947). For example, these theories assume that all decisions under risk or uncertainty can be represented as gambles, with monetary amounts representing the outcomes. In five experiments, we pursue an
opposing, not previously investigated idea: Risky choices are affected by decision content, even when utilities and probabilities are known. We studied the influences of context (Experiments 1 and 2), memory (Experiments 3 and 4) and computational complexity (Experiment 5). In contrast with traditional theories, we find that people do not have underlying
generic preferences; people’s experiences “leak” into decisions even when risk information is explicitly provided.