Independent, not irrelevant: Trial order causes systematic misestimation of economic choice traits.

In fields spanning policy, medicine, and finance, it is increasingly common to guide individual interventions on measures of a person's decision-making characteristics. As the practical application of these measures grows, so have efforts to improve their robustness to ostensibly irrelevant contextual factors such as time of day or ephemeral motivational state. Here, we examine whether such instruments exhibit a fundamental context-dependence: namely, the order in which decision problems are presented. In three datasets evaluating decision-making under different forms of uncertainty, we find systematic, meaningful effects of trial order, which in many cases qualitatively change the measurement’s psychological interpretation (e.g. from risk-seeking to risk-averse). We further examine how trial properties modulate this phenomenon and provide an augmented modeling framework to reliably characterize and correct for these effects.