Journal of Behavioral Decision Making最新文献

Some decision outcomes consist of sequences of single experiences. The aim is to propose a conceptualization of how such sequences are evaluated if affective evaluations of single experiences evoke transient emotional responses with lasting changes in current mood. The conceptualization implies three modes in which the sequences of single experiences are evaluated: (i) Aggregation of affective evaluations of the single experiences retrieved from memory; (ii) Aggregation of current moods associated with emotional responses to the single experiences retrieved or reconstructed from memory; and (iii) Updating of current mood. Simulations of parametrized models are used to compare the different evaluation modes to each other and to show to which extent the simulation results are consistent with some common findings in previous research. The previous research has primarily investigated different rules for aggregation of affective evaluations of single experiences. The simulation results motivate research comparing this mode to the other proposed modes.

This study examines the dynamics of human reliance on algorithmic advice in a situation with strategic interaction. Participants played the strategic game of Rock–Paper–Scissors (RPS) under various conditions, receiving algorithmic decision support while facing human or algorithmic opponents. Results indicate that participants often underutilize algorithmic recommendations, particularly after early errors, but increasingly rely on the algorithm following successful early predictions. This behavior demonstrates a sensitivity to decision outcomes, with asymmetry: rejecting advice consistently reinforces rejecting advice again while accepting advice leads to varied reactions based on outcomes. We also investigate how personal characteristics, such as algorithm familiarity and domain experience, influence reliance on algorithmic advice. Both factors positively correlate with increased reliance, and algorithm familiarity significantly moderates the relationship between outcome feedback and reliance. Facing an algorithmic opponent increases advice rejection frequencies, and the determinants of trust and interaction dynamics differ from those with human opponents. Our findings enhance the understanding of algorithm aversion and reliance on AI, suggesting that increasing familiarity with algorithms can improve their integration into decision-making processes.

Understanding risk tolerance is crucial for predicting and changing behavior across various domains, including health and safety, finance, and ethics. This remains true during a crisis, such as the COVID-19 pandemic, and leads to a key question: Do current risk measures reliably predict risk-taking in the drastically different context of a pandemic? The Domain Specific Risk-Taking (DOSPERT) scale, one of the most widely used risk-taking measures, assesses self-reported risk-taking in response to 30 risky situations across five domains. With the hypothetical risks of the DOSPERT being based on prepandemic circumstances, we estimate that three out of four of its risk-taking situations were not possible due to preventive measures or did not reflect risk-taking in times of COVID-19. In addition, COVID-19 brought forth new behaviors deemed risky. With an aim to better predict risk-taking in times of a pandemic, we introduce the Pandemic DOSPERT (PDOSPERT). We summarize three preregistered online studies with 1254 UK participants to validate the scale against the original DOSPERT and three other common risk-taking measures. We also test its ability to predict pandemic risk-related behaviors at three points in time over 2 years. Overall, we find that the PDOSPERT scale significantly improves predictions for pandemic-related risk behavior as compared to the original DOSPERT. In particular, the health/safety subscale is significantly and strongly associated with pandemic-related risk behavior. We not only validate a pandemic-specific risk task but also introduce a template for developing context- and domain-sensitive measures for risk-taking in the future.

Weijers, R, Wachner, J and Koning, B (2024), The Effect of a Default Nudge on Experienced and Expected Autonomy: A Field Study on Food Donation. J Behav Dec Making, 37: e2404. https://doi.org/10.1002/bdm.2404

For the Acknowledgments, the following was not included, due to the blinding of the manuscript, which was not reverted by the authors when checking the author proofs:

The Acknowledgments should read as follows:

We would like to thank Marleen Lagendijk, Robin Luiten, Harry Wijsbroek, Sarah Puijk, Naomi Markus, Celeste Seijerlin, Eva Buiskool, Fay Olde Dubbelink, and Irissa Pape for their help with data collection.

We apologize for this error.

Past research has shown that people are inconsistent when making predictions about emotions and moral behaviors following their own wrongdoing. However, it is less clear how people react when they did not cause the wrongdoing themselves but the group or collective they associate with. The present paper investigates people's reactions to collective wrongdoing and the question (1) whether the prediction of the experience of group-based guilt is related to actual moral behaviors and (2) whether this prediction is reliable. In three studies, we analyze collective guilt and subsequent behavioral reactions. Study 1 involved real academic situations, varying the kind of unfair treatment of others. A priori, participants overestimated their own subsequent experiences of collective guilt as well as their moral behavior. With respect to actual responses, experienced guilt was the strongest predictor of behavioral reactions, while imagined guilt, in-group identification and satisfaction did not significantly predict responses. Moreover, participants also reacted more to the direct harm caused by their group to others than to unjustified privileges granted to others. Study 2 fully replicated these results and showed relative stability in the predictions of collective guilt. Study 3 compared the responses by participants of the previous two studies with their responses 5 years later, indicating high stability of the observed effects over time. Also, we observed that making repeated predictions after experiencing the guilt-eliciting situation did not improve the accuracy of our participants' predictions. We discuss the implications of these findings for self-predictions, behavioral and affective forecasting of collective emotions, and for common assessment methods of guilt by hypothetical vignettes.

Past research on decisions under risk has documented two contradictory context effects: the contrast effect, where risk preferences in “target” tasks diverge from those in previous “surrounding” tasks, and the assimilation effect that implies the opposite bias. We present four web experiments (three preregistered) that clarify the conditions determining the relative prominence of these opposing effects. Our experiments focus on choice patterns in “target” tasks where participants choose between the status quo and a risky mixed gamble with an expected value of zero. Study 1 examines the impact of surroundings that differ from the target task with respect to the expected benefit from risk-taking. The findings reveal a strong contrast effect: Decreasing the attractiveness of risk-taking in the surrounding tasks increased the risk-taking rate in the target tasks from 53.2% to 79.7%. Study 2 investigates the impact of surroundings that differ in the payoff domain. The findings indicate a strong assimilation effect: Decreasing the attractiveness of risk-taking in the surrounding tasks decreased the risk-taking rate in the target tasks from 74.7% to 36.5%. Additionally, the results revealed unpredicted and robust reversed loss aversion patterns which Studies 3 and 4 further clarify. Our findings (1) suggest that the isolated within-task computations assumed by leading descriptive models overlook substantial contextual considerations, (2) clarify the factors determining the impact of the contrast and assimilation effects in decisions under risk, and (3) provide a theoretical framework for making useful predictions in various scenarios.

This paper examines whether risk preferences in the NBA are reference-dependent and attempts to identify the reference point. Using data from 10 NBA seasons (12,890 games), I find that teams are more likely to attempt a riskier three-point shot (vs. a less risky two-point shot) when below the reference point than above it, consistent with Prospect Theory. The results further show that teams are not influenced by a single fixed reference point, but instead, their choices depend on the score difference, most recent score change, and pregame expectations. Additionally, the weight given to the reference point changes over the course of the game. Teams show a breakeven effect, such that they are more likely to attempt a three-point shot when doing so can tie the game. They also show behavior consistent with mental accounting, as the reference point carries more weight at the end of a quarter than at the beginning. These results provide further real-world evidence for reference-dependent risk preferences while highlighting the challenge of applying reference-dependent models to real-world settings.