Olsson,H.
Juslin,P.
Winman,A.

Comments on an article by Edgar C. Merkle, Winston R. Sieck and Trisha Van Zandt (see record rid]2008-15224-006/rid]). The authors report simulations investigating the effects of random error correction of over-/ underconfidence scores when the assumptions behind these correction procedures are violated. The simulations show that when the underlying process is inconsistent with the assumptions of the correction procedure, for example, if the process involves true biases but the correction procedure presumes that all deviations are random, the correction procedure may overestimate the random error variance and suggest a too large correction. Contrary to what is implied by Merkle et al. we do not view random error as having an "elevated position over other explanations for overconfidence". Random error is only "elevated" in the sense that it is hard to deny that it is a pervasive property of confidence judgments. To the extent that overconfidence and hard-easy effects can be explained as side-effects of random errors of the magnitude empirically observed in data, this weakens the support for models whose main or even sole support is their ability to account for these effects with hypothetical cognitive mechanisms. Neither do we endorse the indiscriminate use of error correction procedures. The applicability of an error correcting procedure can only be established by thorough analyzes of tasks and data, and even then it might be hard to ascertain if the correction is appropriate. (PsycINFO Database Record (c) 2012 APA, all rights reserved)

Olsson, H., Juslin, P., & Winman, A. (2008). The role of random error in confidence judgment: Reply to Merkle, Sieck, and Van Zandt (2008). Journal of Behavioral Decision Making, 21, 449-452. (Full text)

Publication year	2008
Document type:	Article
Publication status:	Published
External URL:	http://dx.doi.org/10.1002/bdm.604 View
Categories:
Keywords: