Fall Research Expo 2021

Debiasing Overconfidence

Individuals and groups often engage in tasks involving predictions or estimates of unknown quantities. 

People often believe that group discussion will surely increase the accuracy of our predictions or estimates, but in fact, group discussion can either help or hurt accuracy. In a recent paper by Silver, Mellers and Tetlock, they discovered a variable called collective calibration which predicts when conversation improves group judgment. Collective calibration means that, prior to discussion, the more knowledgeable members of the group are also more confident and less knowledgeable members are less confident. When groups are collectively calibrated, discussion is more likely to be beneficial.

My research this summer has examined the question of we can get groups to become better collectively calibrated. We know that people, even experts, are often overconfident (e.g., Russo & Schoemaker, 1992). 

Therefore, to help groups improve collective calibration, we can attempt to reduce individual confidence, and we can do so in three different ways: 1) reversing the confidence scale, 2) educating participants about overconfidence, and 3) specifying different conversation goals that could potentially increase or decrease overconfidence.

PRESENTED BY
Other
College of Arts & Sciences 2022
Advised By
Dr. Barbara Mellers
Join Feiyi for a virtual discussion
PRESENTED BY
Other
Velay Fellows Program
College of Arts & Sciences 2022
Advised By
Dr. Barbara Mellers

Comments

As college students who are constantly surrounded by peers, this research is super relevant and I found it interesting.