This
post is by Predrag Teovanović (pictured above), graduate
student at the University of Belgrade. In this post he summarises his recent
paper ‘Individual Differences in Cognitive Biases: Evidence Against One-Factor Theory of Rationality’, co-authored with Goran Knežević and Lazar Stankov,
published in Intelligence.
If
there is a minimal definition of rational behavior, it can be found here.
From the normative standpoint, rational
behaviour is hard (if not impossible) to maintain all the time. Hence, we satisfice by trying to optimize the boundaries of bounded rationality at the
intersection of our own resources (time, information, money, and cognitive
capacities) and environmental demands. Cognitive biases (CBs) emerge in that
junction.
Since
what defines rational behaviour depends on both
environment and organism, and since specific CBs arise in different environments
- it is reasonable not to expect from CBs to be highly related to individual differences in organisms’ capacities and habits. In my article, I
write-up the most forthright empirical demonstration of this expectation that I
could obtain with the great assistance of my dear colleagues Goran Knežević and Lazar Stankov.
I took an approach initiated
by Keith E. Stanovich and Richard F. West (1998,
2000)
and pursued it in a more psychometrical
way. First, I developed new measurement procedures for the assessment of
individual differences in seven CBs (anchoring effect, belief bias,
overconfidence bias, hindsight bias, base rate neglect, outcome bias, and sunk
cost effect). These procedures were devised from conceptual definitions of
aforementioned CB phenomena and/or seminal examples of CB tasks (which are
their operationalisations). On average, dozens of items were used for each CB
phenomenon.
Covariations between items within specific CB
phenomena on a sample of 235 post-adolescents were high enough to produce
satisfactory reliable measures of individual differences (Cronbach’s alphas
were above .70). Thus, it was reasonable to conclude that measurement errors
did not play a major role in this study.
The study was truly aimed to examine a latent
structure of rationality-related phenomena in terms of a relatively
comprehensive list of individual-differences constructs. Results were as
follows: although individual CBs were really relatively independent from one
another (correlations among them were below .25), they were loaded together on two
latent factors that were clearly separated from the latent factor which was combination
of Openness and Need for Cognition (and was labelled as thinking dispositions factor), and well-known psychometric factors of
fluid intelligence (common variance
in Raven, mental rotation and working memory) and crystalized intelligence (vocabulary and verbal analogies).
In short, evidence that a major part of the reliable variance
of cognitive bias tasks was unique implies that a one-factor model of rational
behaviour is not plausible, which further suggests that implying existence of strong
unidimensional underlying construct of rational behaviour is not the right thing to
do. And it was certainly not what Stanovich and West did.