Do Human Reinforcement Learning Models Account for Key Experimental Choice Patterns in the Iowa Gambling Task? DOI Creative Commons

Sherwin Nedaei Janbesaraei,

Amir Hosein Hadian Rasanan, Vahid Nejati

et al.

Computational Brain & Behavior, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 7, 2024

Abstract The Iowa gambling task (IGT) is widely used to study risky decision-making and learning from rewards punishments. Although numerous cognitive models have been developed using reinforcement frameworks investigate the processes underlying IGT, no single model has consistently identified as superior, largely due overlooked importance of flexibility in capturing choice patterns. This examines whether human adequately capture key experimental patterns observed IGT data. Using simulation parameter space partitioning (PSP) methods, we explored two recently introduced models—Outcome-Representation Learning Value plus Sequential Exploration—alongside four traditional models. PSP, a global analysis method, investigates what are relevant parameters’ spaces model, thereby providing insights into flexibility. PSP revealed varying potentials among candidate generate suggesting that selection may be dependent on specific present given dataset. We investigated central fitted all by analyzing comprehensive data pool ( N = 1428) comprising 45 behavioral datasets both healthy clinical populations. Applying Akaike Bayesian information criteria, found Exploration outperformed others its balanced potential experimentally These findings suggested search for suitable reached conclusion, emphasizing aligning model’s with achieving high accuracy modeling.

Language: Английский

Multi-dimensional scaling DOI

Sherwin Nedaei Janbesaraei,

Amir Hosein Hadian Rasanan,

Mohammad Mahdi Moayeri

et al.

Elsevier eBooks, Journal Year: 2025, Volume and Issue: unknown, P. 157 - 186

Published: Jan. 1, 2025

Citations

4
Goodness-of-fit Tests for Categorical Models of Psychological Processes: Fixing the Occasional Failures of Asymptotic Theory DOI
Miguel A. Garcı́a-Pérez, Rocío Alcalá‐Quintana

The Spanish Journal of Psychology, Journal Year: 2025, Volume and Issue: 28

Published: Jan. 1, 2025

Abstract The goodness of fit categorical models psychological processes is often assessed with the log-likelihood ratio statistic ( G 2 ), but its underlying asymptotic theory known to have limited empirical validity. We use examples from scenario fitting psychometric functions psychophysical discrimination data show that two factors are responsible for occasional discrepancies between actual and distributions . One them eventuality very small expected counts, by which number degrees freedom should be computed as J− 1) × I−P−K 0.06 , where J response categories in task, I comparison levels, P free parameters fitted model, K cells implied table counts do not exceed 0.06. second factor administration numbers n i trials at each level x (1 ≤ ). These ridiculously (i.e., lower than 10) they need identical across levels. In practice, when varies it suffices overall N exceeds 40 if = or 50 3, no 10. Correcting using large easy implement practice. precautions ensure validity goodness-of-fit tests based on

Language: Английский

Citations

0
Different exploration strategies along the autism spectrum: Diverging effects of autism diagnosis and autism traits DOI Open Access
Fien Goetmaeckers, Judith Goris, Jan R. Wiersema

et al.

Published: Aug. 16, 2024

When faced with many options to choose from, humans and other agents typically need explore the utility of new choice options. People an autism diagnosis or elevated traits are thought avoid exploring such unknown In a large sample (N = 588), we investigated impact on exploration behavior during value-based decision-making in vast decision spaces. Our findings show that participants were less likely novel more exploit known high-value Computational modeling suggests they engaged uncertainty-driven but exhibited equal random generalization strategies. Interestingly, among non-diagnosed participants, people did not less. highlight important differences strategies between clinical subclinical populations.

Language: Английский

Citations

0
Different exploration strategies along the autism spectrum: Diverging effects of autism diagnosis and autism traits DOI Creative Commons
Fien Goetmaeckers, Judith Goris, Jan R. Wiersema

et al.

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 30, 2024

Abstract When faced with many options to choose from, humans and other agents typically need explore the utility of new choice options. People an autism diagnosis or elevated traits are thought avoid exploring such unknown In a large sample (N = 588), we investigated impact on exploration behavior during value-based decision-making in vast decision spaces. Our findings show that participants were less likely novel more exploit known high-value Computational modeling suggests they engaged uncertainty-driven but exhibited equal random generalization strategies. Interestingly, among non-diagnosed participants, people did not less. highlight important differences strategies between clinical subclinical populations.

Language: Английский

Citations

0
Do Human Reinforcement Learning Models Account for Key Experimental Choice Patterns in the Iowa Gambling Task? DOI Creative Commons

Sherwin Nedaei Janbesaraei,

Amir Hosein Hadian Rasanan, Vahid Nejati

et al.

Computational Brain & Behavior, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 7, 2024

Abstract The Iowa gambling task (IGT) is widely used to study risky decision-making and learning from rewards punishments. Although numerous cognitive models have been developed using reinforcement frameworks investigate the processes underlying IGT, no single model has consistently identified as superior, largely due overlooked importance of flexibility in capturing choice patterns. This examines whether human adequately capture key experimental patterns observed IGT data. Using simulation parameter space partitioning (PSP) methods, we explored two recently introduced models—Outcome-Representation Learning Value plus Sequential Exploration—alongside four traditional models. PSP, a global analysis method, investigates what are relevant parameters’ spaces model, thereby providing insights into flexibility. PSP revealed varying potentials among candidate generate suggesting that selection may be dependent on specific present given dataset. We investigated central fitted all by analyzing comprehensive data pool ( N = 1428) comprising 45 behavioral datasets both healthy clinical populations. Applying Akaike Bayesian information criteria, found Exploration outperformed others its balanced potential experimentally These findings suggested search for suitable reached conclusion, emphasizing aligning model’s with achieving high accuracy modeling.

Language: Английский

Citations

0