A retrieval-based approach to eliminating hindsight bias

ABSTRACT

Individuals exhibit hindsight bias when they are unable to recall their original responses to novel questions after correct answers are provided to them. Prior studies have eliminated hindsight bias by modifying the conditions under which original judgments or correct answers are encoded. Here, we explored whether hindsight bias can be eliminated by manipulating the conditions that hold at retrieval. Our retrieval-based approach predicts that if the conditions at retrieval enable sufficient discrimination of memory representations of original judgments from memory representations of correct answers, then hindsight bias will be reduced or eliminated. Experiment 1 used the standard memory design to replicate the hindsight bias effect in middle-school students. Experiments 2 and 3 modified the retrieval phase of this design, instructing participants beforehand that they would be recalling both their original judgments and the correct answers. As predicted, this enabled participants to form compound retrieval cues that discriminated original judgment traces from correct answer traces, and eliminated hindsight bias. Experiment 4 found that when participants were not instructed beforehand that they would be making both recalls, they did not form discriminating retrieval cues, and hindsight bias returned. These experiments delineate the retrieval conditions that produce—and fail to produce—hindsight bias.

KEYWORDS:

• Hindsight bias
• retrieval-based approach
• SARA
• RAFT
• long-term working memory

Acknowledgements

We thank the classroom teacher for his help in vetting the materials and implementing the experiments. We thank Purav J. Patel for his comments on an earlier version of this article. We thank Jenifer Doll, Tayler Loiselle, Kasey Michel, and Alyssa Worley for their help collecting the data for Experiment 4. Finally, we thank the editor and anonymous reviewers for their thoughtful comments, particularly for their suggestion to run Experiment 4.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Another reason to predict a positive correlation is that the defining differences of the CAI (4) and the HBI (1) contain a common term, .

2. The different effective sizes may also reflect a subtle difference in the procedure of the two experiments. In Experiment 1 (and also Experiments 2 and 3), the filler task between Phase 1 (when participants made their original judgments) and Phase 2 (when they were provided with the correct answers) was to experience a standard science lesson delivered by the classroom teacher. In Experiment 4, the classroom teacher asked that during the filler time, the experimenters introduce themselves and describe how learning about science in middle and high school led them to study science in college and graduate school. It is possible that this more informal filler activity was less effective at clearing the students’ memories for their original judgments, moderating the effect size.

Additional information

Funding

This work was supported by a University of Minnesota Grant-in-Aid awarded to Keisha Varma.