Confidence Flashcards
What is confidence? How accurate are confidence judgements generally?
Self evaluation of ability to do a task. People tend to be overconfident in ability compared to others and are over precise in their estimates, but can sometimes be highly calibrated.
Explain studies showing overconfidence
Dunning Kruger effect, Kiel (2003) illusion of explanatory depth, Svenson (1981) most think better than average driver, Zell et al. review found think better than average across many domains.
Explain studies showing over precise estimates
Hynes and Vanmarcke (1976) - expert engineers estimated height at which structure would break, gave incorrect precise estimates and too narrow confidence interval (range confident would be in) that real answer fell outside.
Ben-David et al. (2013) - financial experts estimated market returns, 80%
confidence intervals only contained answer 36% of time
Explain Koriat 2011. What about Koriat 2008?
Asked which of lines would be longer when stretched out, found greater confidence with greater accuracy for most items - supports that confidence judgements in perceptual decisions are accurate. But found some items people get systematically/consensually wrong, for which confidence negatively correlated with accuracy - supports that confidence judgements in perceptual judgements aren’t accurate. Overall mixed support for accuracy. Koriat 2008 found consensually wrong items for memory (general knowledge questions)
Explain See et al. 2011
Measured self rated power and confidence and other rated willingness to take advice. Found more power correlated with more confident and less likely to take advice. Priming power (with writing passage on time had power versus lacked power) increased confidence and reduced advice taking and performance. Shows that confidence judgements for perceptual decisions aren’t accurate. Shows that confidence judgements are made using actual or perceived power. Shows that confidence is used to decide whether need to take advice or not.
Explain Koriat and Goldsmith 1996
Tested recall and recognition of answers to general knowledge questions,found confidence ratings predictive of memory performance - when given choice to try and recall or pass, more correct answers when choose to recall (82%) compared to (47%) when no choice. Supports that confidence judgements for memory are accurate as using them improves memory. Shows that confidence judgements for memory are used to guide whether attempt recall or not.
Explain Patel et al. 2012
Others are percieves as more confident if their judgements are quicker. Shows that time heuristic is used to judge the confidence of others, could support that use same heuristic to judge confidence of self. Applies to both memory and perceptual decisions.
Explain the evidence theories of how confidence is judged and give some evidence
Vickers and Packer (1982) - race model, accumulate evidence, choose option first meets threshold, confidence from difference in amount of evidence for each option, aligns with neural evidence of build up in activity as making decision.
Pleskac and Busemeyer 2010 - relative evidence accumulation with post decision processing, confidence comes from whether Post decision evidence aligns with decision or not
Boldt et al. 2017 found that evidence variability (evidence less consistently favouring one judgement or another) reduces confidence, independently of how quickly people make decisions (against time heuristic)
Explain the time heustistic theory of how confidence is judged and give supporting evidence
Idea that if make judgement quickly can use as a cue that more likely to be correct.
Costermans et al. 1992 found time spent trying to retrieve is lower with greater confidence (memory).
Kiani et al. 2014 found faster reaction time leads to greater confidence - when making decisions with equal evidence, time predicts confidence (perceptual). Also could be used to judge confidence of others (Patel et al. 2012)
Explain Perfect et al. 2000
Practice with feedback improved accuracy of confidence judgements, feedback in groups makes individuals more calibrated with eachother. Shows that accuracy of confidence can be improved for individuals and groups.
Give evidence against the time heuristic theory for confidence
Vickers and Packer 1982 - people are confident when slow - when asked to perform task slowly/carefully people are more confident and more accurate
Petrusic and Baranski 2009 - confidence predicts accuracy better than reaction times
Explain the confidence heuristic and give evidence
People treat confidence of others as an important cue to their accuracy
Cutler et al. 1988 - mock jury experiment where varied confidence of witness when identifying accused as perpetrator, confident witnesses influence juries more (80% confident led to 39% guilty verdicts and 100% confident led to 64% guilty verdicts), greater than effect of factors known to affect accuracy of memory (whether suspect in disguise, whether delay between event and identification 2 days or 2 weeks) which didn’t have much effect
Explain Tenney et al 2007.
Read about eye witnesses identifying perpetrator, varied confidence of witness, generally greater credibility and guilty rating for high confidence versus low confidence. Shows that judgement of others confidence used to assess accuracy of others. Found that when witness shown to be wrong or give incorrect detail, confident witness got lower judgements of credibility and less guilty judgements, credibility of witness with low confidence not much affected by whether wrong or not but more guilty judgements. Shows that use of confidence of others to assess accuracy is more so when calibrated/sensitive to misplaced confidence.
Explain Bahrami et al. 2010
Confidence has an important role in group decisions as can arbitrate between disagreements by taking answer of most confident member, if calibrated most confident person more likely to be right so improves accuracy, group generally outperforms best performing individual.
But fails when big difference in ability between members as confidence judgements break down Dunning kruger style, trake judgements of worse performing individual.
Shows that confidence used in group decisions and accuracy depends on calibration of group.
Explain Henmon 1911
Asked to say which line is longer and say a-d indicating confidence, found high calibration of percentage correct and confidence. Supports that confidence for perceptual decisions is accurate. Also found that reaction time varies for a fixed confidence - confidence calibrated with accuracy but not reaction time. Is evidence against time heuristic.
