In risk-research, there are two traditions of measurement: the attribute-based and the vignette-based tradition. The attribute- based approach focuses on the impact that the attributes (prob- abilities and outcomes) of risky options have on the processing of risk-related information. The vignette-based approach fo- cuses on responses to questions about contextualized situations involving risk. We bring these two approaches together here to investigate the stability of risk preferences and information processing in risky choice tasks across different contextualized situations. To this end, we employ an evidence-based multi- attribute gamified risky choice task in a retest design. The re- sults (N = 226) show that risk propensities are very stable within domains across time. Participants’ explicit beliefs about risks and returns did not accurately reflect the actual rank order of the costs and benefits of actions in the real world, which we obtained from statistical databases. Also, we find that that pro- spect theory’s risk-attitude parameters are mostly unrelated to the risk-taking in the contextualized task, and that benefit per- ceptions influence risk-taking, in line with a risk-return trade- off view on risk-taking.

Decision making under risk is often studied as a preferential choice governed by stable individual personality characteristics, but risky choice can also be viewed as a dynamic problem of resource accumulation to survive. When decision makers aim to reach a particular goal in limited time, such as “earn at least $100 in five choices,” risky choice becomes a non-trivial planning problem. This problem has an optimal solution that can differ from immediate expected-value maximization. We studied the optimality of risky choices under such minimum goal requirements experimentally and find that the observed choices under goals approximate the optimal solution. However, because the optimal model is very complex, we examine if simpler models can predict people’s choices better. We test an extended version of prospect theory, assuming a dynamic reference point that depends on the distance to the goal. This “dynamic prospect theory” was better than the alternative model in describing people’s decisions (i.e., for 63% of the participants, it was the best model). Our findings show that humans can excel in a highly complex, dynamic, risky choice problem and that a dynamic version of prospect theory provides one possible explanation for how people decide under risk when long-term goals matter.