Modeling risk perception and risk taking in dynamic situations
Final Report Abstract
This research project P5 was part of the DFG Research Unit RiskDynamics (FOR 2374) at the University of Konstanz. Its goal was to investigate and model risk perception as well as risk taking in dynamic situations. Originally, our perspective on dynamic situations concerned between and within individual decisions. For both aspects, we planned experiments and computationally modeling. However, early in the project it became clear that focusing on between-decision dynamics was more promising. Moreover, stronger synergies across some projects of the Research Unit could be expected. Therefore, we conducted various experiments to examine how people deal with the dynamic nature of risky decision making. A key task for this endeavor was the Columbia Card Task (CCT). In first experiments it was tested to what extent the CCT relates to other measures of risk, and aspects of real-life risk taking. It was also examined whether shorter versions of the CCT show a similar relationship. We found that CCT scores did not generally match with other frequently used measures of risk-taking. Merely recreational risk-taking was weakly related with the 32-card CCT score. In a further experiment, we investigated to what extent subjective control in the CCT affects the degree of individual risk taking. Although control had a significant effect, the effect size was rather small. We had more success in a series of experiments, where we investigated whether sequential decisions were either based on conditional probabilities or on conjunctive probabilities. Our findings show that the performance can best be accounted for by assuming that people rely on a mixture of conditional and conjunctive probabilities. Originally, we planned to use drift diffusion models for modeling the dynamics of individual choice behavior. Unfortunately, this did not work out. Therefore, we applied other models to our data and to data from other projects in the research unit. In collaboration with colleagues from P4, we examined the influence of mindsets on risk taking in the CCT. We compared several stationary and non-stationary models. Unfortunately, modeling did not add insights beyond behavioral analyses. In contrast, modeling of the data from project P6 was successful. We were able to identify model parameters that reflect decision deficits in people suffering from alcohol dependence. Taken together, much effort was investigated in this project and in collaborations with other projects from the research unit. Unfortunately, the success was only moderate. Because the usefulness of the CCT turned out to be rather limited, and, therefore, our main ideas could only partly be realized, we did not apply for an extension of the project.
Publications
- (2020). Are choices based on conditional or conjunctive probabilities in a sequential risk-taking task? Journal of Behavioral Decision Making, 33(3), 333-347
Haffke, P., & Hübner, R.
(See online at https://doi.org/10.1002/bdm.2161) - (2020). Journal of Addictive Diseases, 39(1), 88-95
Senn, S., Odenwald, M., Sehrig, S., Haffke, P., Rockstroh, B., Pereyra Kröll, D., Menning, H., Wieber, F., Volken, T., & Rösner, S.
(See online at https://doi.org/10.1080/10550887.2020.1820810)
