The construct-behavior gap and the description-experience gap: Comment on Regenwetter and Robinson (2017).

Regenwetter and Robinson (2017) discuss a challenging construct-behavior gap in psychological research. It can emerge when testing hypotheses that pertain to a theoretical construct (e.g., preferences) on the basis of observed behavior (e.g., actual choices). The problem is that the different heuristic methods that are sometimes used to link overt choices to covert preferences may ignore heterogeneity between and within individuals, rendering inferences drawn from choices to preferences invalid. Regenwetter and Robinson's remedy is to make heterogeneity an explicit part of the theory. They illustrate the problem and a remedy to it with the description-experience gap (D-E gap), the systematic gap in choices based on described versus 'experienced' probabilities. We welcome their sophisticated reanalysis of some early data sets, which, by taking heterogeneity into account, finds strong evidence for a D-E gap in probability weighting. Yet we see three issues with the remedy, which we likewise highlight using the D-E gap. First, the D-E gap cannot be reduced solely to probability weighting but rather unfolds across several different psychological constructs suggesting that part of the construct-behavior gap may stem from trying to reduce multidimensional behavior to a single construct. Second, the authors' modeling of heterogeneity leaves aside the heterogeneity of people's sampled experience in decisions from experience, which highlights the importance of also considering the potential causes of heterogeneity. Third, we identify potential sources of heterogeneity in choice behavior that go beyond probabilistic responses and preferences and advocate for a pluralistic approach to modeling it. Last but not least, we emphasize that, notwithstanding the importance of rigor and logical coherence in scientific theories, simplifications and (false) generalizations are indispensable in the pursuit of scientific knowledge. (PsycINFO Database Record (c) 2018 APA, all rights reserved).