The Curious Complexity between Confidence and Accuracy in Reports from Memory

Abstract

Th e relation between the probability of remembering an event and one's confi dence in it seems obvious: Th e more confi dent a person is in remembering an event, the more accurate he or she will be (and vice versa). Imagine giving people a series of events to remember every day for a week, say 10 per day. Th e events could be sentences such as "Th e hippie touched the debutante in the park" or "Th e policeman arrested the homeless woman near the movie theater." Th en, on the seventh day, people could be asked to recall (or recognize) all the sentences that had been presented that seventh day, and those from the third day of the experiment, and to rate the confi dence of each reported memory. It would surprise no one to learn that people would correctly remember more sentences from the seventh day than from the third day; surely they would be more accurate for the recent memories. In addition, there is no doubt that their confi dence would track their accuracy if confi dence were measured on, say, a 7-point rating scale (from 7 = sure the event happened to 1 = sure the event did not happen). People would be much more confi dent for the recently presented sentences than for those heard 4 days previously. Th e reason people can intuit the result of this experiment so accurately is that we essentially live this experiment every day of our lives. We can tell an inquirer what events happened to us today with reasonable accuracy and certainty, but if we are asked to retrieve events from a particular day even a few days ago, they would be much hazier to us-we would be less accurate and less confi dent.

Receiver operating characteristic analysis of age-related changes in lineup performance

“…First, one particularly important issue is that confidence should be predictive of accuracy. In the adult eyewitness literature, confidence, if it is taken immediately after the lineup identification decision, is predictive of accuracy (see Mickes et al, 2012, andRoediger, Wixted, &DeSoto, 2012, for a review). The evidence is mixed in children, however, with some studies indicating that confidence can be predictive of accuracy (e.g., Leippe et al, 1991;Parker & Carranza, 1989;Parker & Ryan, 1993), yet others finding that children's confidence is not predictive of accuracy (e.g., Brewer & Day, 2005;Keast et al, 2007).…”

“…Research using an alternative method, the calibration approach, which assesses whether confidence ratings provide a probabilistic estimate of accuracy, children's meta-cognitive judgments have also been found to be less accurate than adults (Keast, Brewer, & Wells, 2007). Yet, for adult witnesses, when confidence is measured immediately following the lineup identification decision, confidence can predict accuracy (see Mickes et al, 2012, andRoediger, Wixted, &DeSoto, 2012, for a review). In view of this literature, we expected to find age-related changes in the relationship between confidence and accuracy, with the association being weaker in children compared to adults.…”

Why eyewitnesses fail

“…An eyewitness may be wrong for the reasons described above, but a witness who testifies in court with confidence is generally very compelling to triers of fact (26). Contrary to common intuition, however, courtroom statements of confidence are very poor predictors of accuracy (26)(27)(28)(29). The cause of this confidence-accuracy disparity is well captured by Daniel Kahneman's cognitive "illusion of validity" (30):…”

Confidence in forced-choice recognition: What underlies the ratings?

“…There are various ways of asking the question of whether confidence tracks accuracy (see Roediger, Wixted, & DeSoto, 2012, for a detailed discussion). One can be interested whether people are more confident in their more accurate responses or whether people who express greater confidence in their responses are actually more accurate.…”