=Paper=
{{Paper
|id=Vol-1419/paper0121
|storemode=property
|title=Individual Differences in Performance on Iowa Gambling Task are Predicted by Tolerance and Intolerance for Uncertainty
|pdfUrl=https://ceur-ws.org/Vol-1419/paper0121.pdf
|volume=Vol-1419
|dblpUrl=https://dblp.org/rec/conf/eapcogsci/KornilovKKC15
}}
==Individual Differences in Performance on Iowa Gambling Task are Predicted by Tolerance and Intolerance for Uncertainty==
https://ceur-ws.org/Vol-1419/paper0121.pdf
Individual Differences in Performance on Iowa Gambling Task are Predicted by
Tolerance and Intolerance for Uncertainty
Sergey A. Kornilov (sa.kornilov@gmail.com)
Department of Psychology, Moscow State University
11/5 Mokhovaya St., Moscow, Russian Federation 01
Evgenii Krasnov (evkrasnov@gmail.com)
Department of Psychology, Moscow State University
11/5 Mokhovaya St., Moscow, Russian Federation 01
Tatiana V. Kornilova (tvkornilova@mail.ru)
Department of Psychology, Moscow State University
11/5 Mokhovaya St., Moscow, Russian Federation 01
Maria A. Chumakova (chumakova.mariya@gmail.com)
Department of Psychology, Moscow City University for Psychology and Education
29 Sretenka St., Moscow, Russian Federation 01
Abstract experimental learning task, IGT is rarely analyzed in terms
Iowa Gambling Task (IGT) is frequently used to index
of participants’ trajectories over time. The study reported in
individual differences in decision-making under uncertainty, this paper aimed to partially address these three gaps in the
particularly in atypical (clinical) populations. However, it is literature by investigating the role of the complex traits of
rarely analyzed as a learning task, and research on the tolerance and intolerance for uncertainty in participant’s
predictors of performance on the IGT in normative learning during decision making under uncertainty in the
populations is scarce. Here, we focused on tolerance and IGT task.
intolerance for uncertainty as two traits that could potentially The IGT requires the participant to choose cards from
influence subjects’ IGT performance. Using mixed modeling
analysis of longitudinal experimental data (n=60, 5 blocks) four decks that have a systematically varied intermittent
we showed that tolerance for uncertainty predicted the initial gain and loss structure that the participants uncover by trial
level of risk in IGT as manifested in the proportion of “bad and error during the experiment. The two disadvantageous
decks” chosen; at the same time, intolerance for uncertainty IGT decks (A and B) are associated with high immediate
predicted explorative learning in IGT as manifested in the rewards but long-term net losses. The two advantageous
number of deck switches after a loss and its decline over the decks (C and D), on the other hand, are associated with
course of the experiment. The results are discussed in the
lower immediate rewards but also significantly smaller
context of viewing IGT as capturing a set of dynamic decision
making processes that rely on learning, risk taking, and long-term losses. Initially used to test the somatic marker
exploration. hypothesis in patients with lesions to the ventromedial
prefrontal cortex (Bechara, Damasio, Damasio, &
Keywords: decision making; learning; Iowa Gambling Task;
Anderson, 1994), IGT has since been productively used
tolerance for uncertainty; intolerance for uncertainty; risk
with clinical (e.g., psychiatric and neurological) as well as
developmental (i.e., adolescents) populations to study
Introduction
decision making.
Perhaps surprisingly, a recent review of the associations
Despite a considerable body of research generated in the between participants’ performance on the IGT task and
field of decision making in the recent several decades, cognitive traits found that IGT performance was largely
sources of individual differences in decision making remain unrelated to general cognitive ability, working memory,
largely understudied, in part due to the absence of well- executive functions, and set shifting, although no aggregate
established individual differences-focused research effect sizes were computed (Toplak, Sorge, Benoit, West, &
paradigms (e.g., see Jackson & Kleitman, 2014). Moreover, Stanovich, 2010). These results highlighted the distinction
when individual differences in decision making become the between cognitive processes captured by the maximum
focus of the investigation, such studies frequently center performance (i.e., intelligence testing) measures and
around cognitive traits (i.e., intelligence and emotional measures of rational decision making. At the same time,
intelligence) and the “traditional” set of Big Five personality participants’ performance on IGT was found to be
traits to explain participants’ performance on tasks like the modulated by trait anxiety and neuroticism (Hooper,
Iowa Gambling Task (IGT). Finally, although inherently an Luciana, Wahlstrom, Conklin, & Yarger, 2008; Miu,
728
Heilman, & Houser, 2008). At the same time, personality We analyzed the following indices of performance on the
correlates of participants’ performance on the IGT task have IGT: 1) cumulative Net Gain, 2) proportion of
rarely been examined in non-clinical samples (Buelow & disadvantageous deck choices (Bad Decks) to total deck
Suhr, 2009). choices, and 3) proportion of deck switches after
Note that IGT performance is most frequently analyzed in experiencing a loss (Loss Switches). Participants’
terms of the resulting proportion of disadvantageous choices performance was averaged across 20 trials within each of
to all choices [(A+B)/(C+D)] in the second half of the the five blocks, and the resulting data were subjected to
experiment (that typically consists of 5 blocks of 20 trials, mixed modeling (or growth curve) analysis (see below).
for a total of 100 trials) or the overall game money net gain
by the end of experiment. Yet, the IGT task can also be
considered to be a learning under uncertainty task where New Questionnaire of Tolerance for Uncertainty
participants are faced with the neccessity to establish and (NTN)
continuously refinine probabilistic representations of the The previously validated New Questionnaire of Tolerance
reward and punishment structure of the environment (i.e., for Uncertainty (NQTU, or NTN in Russian) was used to
the experimental deck setup). Correspondingly, decision measure variables associated with acceptance of uncertainty
making in IGT unfolds over time and within-participant (Kornilova, 2010). This self-report questionnaire showed
learning trajectories can be established and related to superior psychometric properties compared to other existing
individual differences in participants’ basic cognitive and measures of the same construct(s). We used the following
personality characteristics. two subscales of the NQTU for the purpose of the study -
Uncovering these trajectories and explaining them from Tolerance for Uncertainty and Intolerance for Uncertainty.
the standpoint of individual differences was the main aim of Tolerance for Uncertainty (TU) was conceptualized as the
the reported study. Based on our previous findings of the readiness to make decisions and act in uncertain situations,
importance of tolerance and intolerance for uncertainty for openness to new ideas, changing stimuli and changing
understanding the nature and mechanisms of decision thinking strategies. In the original structural equation model
making, we hypothesized that longitudinal indices of IGT (SEM) reported in Kornilova’s (2010) study, TU was one of
performance should be related to traits of tolerance and the indicators of the latent variable of acceptance of
intolerance for uncertainty as capturing the fundamental uncertainty and risk (which also included
regulatory elements of decision making (Chumakova & experiential/intuitive thinking style). In this model,
Kornilov, 2013; Kornilova, 2013). tolerance for uncertainty was a construct relatively
independent of intolerance for uncertainty. Intolerance for
Uncertainty (ITU) was conceptualized as willingness to
Methods achieve clarity in the world (including the world of ideas),
rejection of uncertainty in judgement, rigidity and
Participants rationality (directed towards acquiring maximum
The participants were undergraduate students from Moscow information required for decision making).
State University and military instructors. A total of 60 adult
participants took part in the study (age ranged from 18 to Results
52, M = 30.58, SD = 10.61; 41 were males, and 19 were
females).
The data were analyzed using a set of mixed linear models
(Baayen, 2008) as implemented in the lme4 R package
(Bates & Maechler, 2010). Net Gain, Bad Decks, and Loss
Iowa Gambling Task (IGT)
Switches were used as dependent variables. Block number,
All participants were first administered the Iowa Gambling sex (0=females, 1=males), and TU/ITU scores were entered
Task, followed by personality assessments. For the purpose in the model as fixed effects. Block number was centered at
of the study, we translated and adapted the standard the value of 1, age and TU/ITU scores were mean-centered.
computerized IGT protocol developed by Grasman and The unconditional growth models also included the
Wagenmakers (Grasman & Wagenmakers, 2005). Briefly, quadratic growth term when appropriate (as determined by a
participants were instructed to choose cards from one of set of comparisons of nested models). In conditional growth
four decks presented on the screen – A (+$100 or -$150, - models, age, sex, and ITU/TU predicted both the intercept
$200, -$250, -$300, -$350 with a probability of 50%); B and the growth parameters. Intercept and growth parameters
(+$100 or -$1,250 with a probability of 10%); C (+$50 or - were also included as random effects in all of the models.
$50 with a probability of 50%); D (+$50 or -$250 with a First, we found that over the course of the experiment,
probability of 10%). The experiment was organized in 5 participants exhibited significant learning that could be
blocks of 20 trials, and feedback was provided after each described by a quadratic function (see Table 1). There was a
trial on the screen of the computer, along with the feedback trend for the association between participant’s net gains for
regarding the participant’s overall progress on the task (i.e., the first IGT block (i.e., the intercept parameter) and TU (B
net gain and losses). = 5.68, SE = 2.93, t = 1.94), suggesting that tolerance for
729
Table 1 : Summary of Fixed Effects for the Mixed Models with IGT Performance as Dependent Variables
Net Gain Bad Decks Loss Switches
Parameter B SE t B SE t B SE t
Intercept 1514.06 52.70 28.73* .53 .05 10.39 3.85 .50 7.72
Block 1557.07 78.10 19.94* -.04 .02 -2.03* -.32 .55 -.58
Sex 104.97 70.63 1.49 .07 .07 1.09 .85 .67 1.27
Age -3.46 2.86 -1.21 -.002 .003 -.90 -.03 .03 -1.22
TU 5.68 2.93 1.94t .006 .003 2.13* -.04 .03 -1.51
ITU 1.25 2.23 .56 .002 .002 .90 -.05 .02 -2.32*
Block2 -26.13 12.45 -2.10* n/a n/a n/a .05 .13 .38
Block:Sex -8.01 104.68 -.08 .02 .03 .93 -.40 .73 -.55
Block:Age .42 4.24 .10 -.001 .001 -.60 .02 .03 .52
Block:TU 6.18 4.34 1.42 -.002 .001 -1.42 .03 .03 1.12
Block:ITU 3.59 3.30 1.09 -.001 .001 -.71 .05 .02 2.14*
Block2:Sex 23.09 16.69 1.38 n/a n/a n/a -.02 .17 -.10
Block2:Age -.63 .68 -.94 n/a n/a n/a -.0002 .007 -.02
Block2:TU -.93 .69 -1.34 n/a n/a n/a -.008 .007 -1.17
Block2:ITU -.56 .52 -1.07 n/a n/a n/a -.01 .005 -1.86t
t
Note. * - p < .05; – p < .10. TU – tolerance for uncertainty; ITU – intolerance for uncertainty
uncertainty modulates the baseline IGT performance level. tolerance for uncertainty. These results suggests that
We also found that the proportion of Bad Decks tolerance for uncertainty regulates baseline risk taking
decreased over the course of the experiment linearly. propensity during online decision making under uncertainty,
Importantly, TU was a significant predictor of the baseline consistent with previous reports of TU being linked to risk
level for this dependent variable (B = .006, SE = .003, t = taking (Kornilova, 2013) and reports of significant
2.13), corroborating results reported in the previous associations between IGT performance and sensation
paragraph. seeking (Suhr & Tsanadis, 2007). Given that we also found
Finally, we found that Loss Switches were relatively a trend for TU being a positive predictor of baseline IGT net
constant over the course of the experiment for our “average” gain, these results suggest that TU indexes processes that
participant. Yet, ITU predicted the baseline level (B = -.05, play important roles in environment sampling and the
SE = .02, t = -2.32), with higher ITU associated with lower development of probabilistic representations (i.e., learning
number of deck switches after losing experimental money. that manifests in disadvantageous decks aversion) at the
ITU also predicted the linear growth parameter (B = .05, SE initial stages of decision making that determine the baseline
= .02, t = 2.14) and showed a trend for a significant “performance corridor”.
association with the quadratic growth parameter as well (B On the other hand, we found that higher intolerance for
= -.01, SE = .005, t = -1.86). This result suggests that uncertainty was associated with lower baseline exploratory
individuals with higher ITU are less likely to explore other activity after failure (i.e., number of deck switches after
decks after losing money in the beginning of the experiencing a monetary loss in the IGT). This finding is
experiment, and potentially, display a relatively constant (or consistent with viewing intolerance for uncertainty as
slightly increasing, compared to constant or slightly indexing risk aversion and uncertainty rejection (Kornilova,
negative average growth rate, see Table 1) level of deck 2010) and recent reports of ITU being associated with
switching throughout the course of the experiment. avoidant behavior in decision making under uncertainty
(Luhmann, Ishida, & Hajcak, 2011).
Discussion Overall, the results of our study indicate that decision-
making under uncertainty is partially modulated and
The reported study investigated individual differences in the regulated by tolerance and intolerance for uncertainty.
dynamic indices of decision making as captured by the IGT. Interestingly, TU appears to regulate baseline risk
We found that tolerance and intolerance for uncertainty propensity that underlies exploratory learning at the initial
predicted several of these indices, suggesting that these stages of decision making, while ITU regulates risk
traits modulate decision making on-line. propensity after failure/loss, potentially constraining
Tolerance for uncertainty predicted the participant’s learning under uncertainty through risk aversion and
baseline performance on the IGT – i.e., individuals with outcome sensitivity.
higher tolerance for uncertainty were more likely to choose Our study was also instrumental in showing the added
disadvantageous decks in the beginning of the experiment, value of investigating dynamic (as opposed to static) indices
and yet showed higher net gains than individuals with lower of decision making in the IGT task. Future studies should
730
investigate the incremental predictive value of TU/ITU with choice and decision making]. Psilkhologicheskii
respect to IGT performance over and above cognitive traits Zhurnal, 34(3), 89-100.
(i.e., nonverbal and verbal intelligence, working memory, Luhmann, C. C., Ishida, K., & Hajcak, G. (2011).
and executive functions) in larger samples and attempt to Intolerance of Uncertainty and Decisions About
clarify the mechanistic role of TU and ITU in constraining Delayed, Probabilistic Rewards. Behavior Therapy,
the online development of probabilistic representations and 42(3), 378-386.
risky exploratory behavior under uncertainty. Miu, A. C., Heilman, R. M., & Houser, D. (2008). Anxiety
impairs decision-making: Psychophysiological
Acknowledgments evidence from an Iowa Gambling Task. Biological
This research and the preparation of this paper were Psychology, 77(3), 353-358.
supported by the Russian Foundation for Humanities Suhr, J. A., & Tsanadis, J. (2007). Affect and personality
(RGNF), grant №15-06-10404a (PI: Smirnov). correlates of the Iowa Gambling Task. Personality
and Individual Differences, 43(1), 27-36.
Toplak, M. E., Sorge, G. B., Benoit, A., West, R. F., &
References Stanovich, K. E. (2010). Decision-making and
cognitive abilities: A review of associations
Baayen, R. H. (2008). Analyzing linguistic data: A practical between Iowa Gambling Task performance,
introduction to statistics. Cambridge: Cambridge executive functions, and intelligence. Clinical
University Press. Psychology Review, 30(5), 562-581.
Bates, D. M., & Maechler, M. (2010). lme4: Linear Mixed-
Effects Models Using S4 Classes. R package.
(Version 1.0-6). Retrieved from http://R-Forge.R-
project.org/projects/lme4/
Bechara, A., Damasio, A. R., Damasio, H., & Anderson, S.
W. (1994). Insensitivity to future consequences
following damage to human prefrontal cortex.
Cognition, 50(1–3), 7-15.
Buelow, M. T., & Suhr, J. A. (2009). Construct Validity of
the Iowa Gambling Task. Neuropsychological
Review, 19, 102-114.
Chumakova, M. A., & Kornilov, S. A. (2013). Individual
differences in attitudes towards uncertainty:
evidence for multiple latent profiles. Psychology in
Russia: State of the Art, 6(4), 94-108.
Grasman, R. P. P. P., & Wagenmakers, E. J. (2005). A
DHTML implementation of the Iowa Gambling
Task. Retrieved from
http://purl.oclc.org/NET/rgrasman/jscript/IowaGa
mblingTask
Hooper, C. J., Luciana, M., Wahlstrom, D., Conklin, H. M.,
& Yarger, R. S. (2008). Personality correlates of
Iowa Gambling Task performance in healthy
adolescents. Personality and Individual
Differences, 44(3), 598-609.
Jackson, S., & Kleitman, S. (2014). Individual differences in
decision-making and confidence: capturing
decision tendencies in a fictitious medical test.
Metacognition and Learning, 9(1), 25-49. doi:
10.1007/s11409-013-9110-y
Kornilova, T. V. (2010). Novii oprosnik tolerantnosti-
intolerantnosti k neopredelennosti [New
Questionnaire for Tolerance and Intolerance for
Uncertainty]. Psikhologicheskii Zhurnal, 30(6),
140-152.
Kornilova, T. V. (2013). Psikhologia neopredelennosti:
Edinstvo intellektualno-lichnostnoi reguliatsii
reshenii i vyborov [Psychology of uncertainty: The
unity of cognitive and personality regulation of
731