NoteFren home

Flashcard library · Psychology · PaperFren

Separating Habit from Value in the Iowa Gambling Task

A new math model shows that separating our habit of repeating past choices from our calculations of future rewards better explains human decision-making.

Keep what you just learned

Save these 35 cards to your NoteFren library. Your first review starts right away, and each card links back to the study it came from.

What is the primary objective given to participants in the Iowa Gambling Task (IGT) experiment described in this study?

Participants are given a starting capital of 2000 points and instructed to maximize gains and minimize losses in order to finish the task with at least 2500 points.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

How does the Delta learning rule differ from the Decay learning rule regarding how unchosen options are updated?

Under the Delta rule, the expected values of unchosen options remain unchanged, whereas the Decay rule assumes that the expected values of all options discount or decay on every trial.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

Why does the Decay learning rule inherently allow computational models to predict perseverative choice behavior?

Since unchosen options decay in value, a frequently chosen option's relative value increases compared to the declining values of other options, making it progressively more likely to be selected again.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What is a major theoretical limitation of the standard Win-Stay-Lose-Shift (WSLS) model of decision-making in the IGT?

It assumes that participants ignore the long-term relative expected values of the options and base their decisions purely on the immediate positive or negative outcome of the previous trial.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What critical shortcoming does the standard Prospect Valence Learning (PVL) model have regarding expected value and perseveration?

The PVL model conflates expected value maximization with the tendency to persevere because it utilizes a single mathematical value to represent both behavior patterns.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What was the main developmental goal of the Value-Plus-Perseveration (VPP) model introduced in this study?

The VPP model was designed to decompose and separately represent the independent tendencies of expected value maximization and perseveration by utilizing distinct mathematical terms.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What does the recency parameter (phi) in the Delta value-updating rule represent?

The recency parameter determines the weight given to the most recent trial outcome when updating the expected value of the chosen option.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

In the VPP model's value integration equation, what does a weight parameter (w_Ej) of less than 0.5 represent?

It indicates that the overall selection probability of an option is influenced more by its perseverative strength than by its expected value.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What are the two free parameters estimated in the WSLS model utilized by the authors?

The parameters are the probability of staying with the same option after a win, and the probability of shifting to a different option after a loss.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What baseline assumption does the Baseline model make regarding option selection in the IGT?

The Baseline model assumes fixed, trial-independent choice probabilities for selecting each deck throughout the task.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

How did participants' behavior change across the blocks of the IGT experiment?

Participants demonstrated learning over the course of the experiment by progressively choosing the advantageous decks more frequently in later blocks.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

Which statistical criteria did the authors use to evaluate and compare the fit of the different computational models to the behavioral data?

They used Akaike's Information Criterion (AIC) and the Bayesian Information Criterion (BIC), both of which penalize models for having a higher number of free parameters.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

When comparing only the single-term models, how did the fits of the Decay rule models compare to the Delta rule models?

The Decay rule models achieved very similar fits to each other, and both provided a superior fit to the data compared to the Delta rule models.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

Between the two models utilizing the Delta learning rule, which utility function provided a better fit to the behavioral data?

The model utilizing the Prospect Valence (PV) utility function fit the experimental data significantly better than the model utilizing the Expectancy Valence (EV) utility function.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

Which of all the tested models provided the overall best fit to the empirical IGT data?

The Value-Plus-Perseveration (VPP) model provided the best fit to the data according to both AIC and BIC.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

How did the authors determine parameter values when generating simulated data sets for each model?

They generated 1000 simulated data sets per model by sampling parameter combinations with replacement from the best-fitting parameter sets of the human participants.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

Why did the single-term Delta rule models over-predict the number of switch trials in simulations compared to human behavior?

Because they lacked any built-in decay or perseverative mechanism, which caused them to under-predict the human tendency to stick with the same option.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

Why did the single-term Decay rule models under-predict the overall number of switch trials in task simulations?

Because they conflate perseverative behavior with expected value maximization, causing them to over-predict perseveration and under-predict switching.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI

How closely did the VPP model's simulated switch trials match the behavioral data of human participants?

The VPP model's simulated switch trials were nearly equivalent to the human average, matching almost perfectly when rounded to the nearest whole number.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

How was the learning rate parameter (phi) in the PVL Delta rule model correlated with task performance, and what does this imply?

It was significantly negatively correlated with performance, implying that paying less attention to immediate outcomes and averaging over longer periods led to better expected value estimation.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What association was found between WSLS model parameters and participant performance on the IGT?

There was a significant positive correlation between the probability of shifting after a loss and overall task performance.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

Why did the single-term Decay models show almost no relationship between their loss aversion parameters and overall performance?

This null relationship occurred because the Decay models conflate representations for expected value maximization and perseverative behavior, obscuring the impact of loss aversion on choices.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What was the relationship between the VPP model's estimated loss aversion parameter (lambda) and task performance?

There was a strong positive association, meaning that higher estimated loss aversion was strongly linked to better performance on the IGT.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What is a known anomaly when estimating individual participant parameters using Maximum Likelihood Estimation (MLE) in these tasks?

MLE often produces extreme parameter estimates that cluster right on the boundaries of the defined parameter space.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

How did the Bayesian hierarchical estimation approach resolve the boundary anomaly of Maximum Likelihood Estimation?

By yielding posterior distributions, the modes of the parameter distributions were pulled away from the extreme boundaries of the parameter space while still maintaining a strong correlation with performance.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

How did the VPP model's loss aversion parameter correlate with specific selections of the disadvantageous decks (Decks A and B)?

It had a negative correlation with selections of both decks, but the association was only statistically significant for Deck B selections.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What does previous literature suggest regarding the short-term versus long-term predictive strengths of Decay learning rules versus Delta learning rules?

Decay learning rules excel at making short-term, trial-to-trial predictions, whereas Delta learning rules excel at making long-term predictions across an entire sequence of choices.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

According to the authors, why do Decay rule models exhibit an advantage in short-term trial prediction?

Because their mathematical structure implicitly accounts for the short-term tendency of human participants to persevere on choices.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI

Why do Delta rule models show an advantage in long-term sequence prediction compared to Decay rule models?

Because they do not conflate expected value representation with perseveration, allowing them to better model stable personal traits like loss aversion.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What do the authors predict regarding the performance of the VPP model under the generalization criterion method?

They predict that because it isolates perseveration from expected value representation, it will perform well in predicting behavior across different decision-making tasks.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

In the VPP model, how is the perseveration term updated differently based on trial outcomes?

The perseveration term is updated using different increment parameters (epsilon_pos versus epsilon_neg) depending on whether the net trial outcome was positive or negative.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

What do positive versus negative values for the VPP model's perseveration increment parameters (epsilon) indicate?

Positive values indicate a tendency to persevere (stay) on subsequent trials, whereas negative values indicate a tendency to switch options.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

How did the VPP model's exploitation parameter (c) relate to behavioral performance?

Higher values of the exploitation parameter (c) were significantly positively associated with better overall performance in the IGT.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

Why is basic behavioral analysis (such as advantageous vs. disadvantageous choices) insufficient for analyzing IGT performance, according to the authors?

Because the IGT is complex, and basic metrics cannot capture underlying cognitive nuances, making computational model-fitting essential to decompose specific decision-making processes.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified

How has the Expectancy Valence (EV) model historically been used to characterize decision-making deficits in clinical populations?

It has been used to identify whether specific clinical groups attend more to gains than losses, losses than gains, or focus primarily on recent outcomes.

Source: Decomposing the roles of perseveration and expected value representation in models of the Iowa gambling task · DOI · quote verified