From Uppsala to Umeň
This report is written in the midst of the turmoil of moving from Uppsala to Umeň, a 600 km drive by car with two very lively six-month babies (twins) in the back-seat.
As you may know, Umeň was once the site of Brunswikians like Berndt Brehmer and Mats Bj÷rkman, and the chair that I have received was indeed occupied by Mats in the 60s and 70s.
Our continued research on probability judgment in Uppsala has resulted in a couple of articles and manuscripts. In a recent article (Juslin, Wennerholm, & Olsson, 1999, JEP:LMC, 25, 1038-1052), we explore the phenomenon of format-dependence, the simultaneous observation of over- and underconfidence when the same judgment tasks are approached with different probability assessment formats.
In another article (Juslin, Winman, & Olsson, in press, Psychological Review) we perform a "meta-analysis" that compares the over/underconfidence for all representative and selected samples of general knowledge items that we could find. Across representative data sets over/underconfidence bias was .01 and across selected data sets over/underconfidence was .09, a difference that is not accountable in terms of the difference in proportion correct. Moreover, we demonstrate that most of the notorious "hard-easy effect" seems to be explained by statistical artifacts (e.g., linear dependency).
The manuscript where we apply the sensory sampling model (Juslin & Olsson, 1997, Psychological Review, 104, 344-366) to distinguish between cue-heuristic, inferential processes from direct-perceptual processes in a task that involves judgments of the relative mass of colliding objects, is now forthcoming in Psychological Review (Runeson, Juslin, & Olsson, in press).
In a more recent experiment, we compare probability judgments in simple sensory discrimination tasks to probability judgments in general knowledge tasks and document a number of additional differences between the two kinds of tasks, for example, in regard to additivity (Juslin, Winman, & Olsson, in preparation). Henrik Olsson will present his doctoral dissertation on the sensory sampling model in December 1999.
Magnus Persson and I are finishing a manuscript that applies the context model by Medin and Schaffer to probability judgment and decision tasks (the modification is entitled PROBEX for probabilities from exemplars). In the manuscript, we explore the ecological rationality (Ó la Gigerenzer and colleagues) of an algorithm that makes judgments by rapid retrieval of a small number of stored exemplars, and evaluates its quantitative fit to performance data for "real" participants performing the same task.
Finally, we have developed a rule-based account of the inverse base-rate effect in categorization studies that traces the effect to eliminative inferences (Juslin, Wennerholm, & Winman, 1999). This explanation, that may have novel normative implications, contrasts with previous accounts that revolve, in one way or another, around the theme of cue-competition (e.g., in the Rescorla-Wagner model).