The Wisconsin Card Sorting task is rather famous but appears to be quite difficult to model computationally.

I work in RL and I am interested in how people learn the optimal strategy. I'm interested in the task because it would allow a number of experimental manipulations. I want to capture the role of memory in the optimal strategy. People can keep track of some, but not very much, history of trials when doing the task. (I.e the mathematically optimal strategy seems too computationally heavy.)

  • How could I approach modelling the Wisconsin Card Sorting task?
  • What other research has modelled the task?
  • 1
    $\begingroup$ What paradigm are you working in, and why does the first result in Google Scholar (Dehaene & Changeux, 1991) not answer your question? Also, what is your reason for believing it is hard to model? $\endgroup$ – Artem Kaznatcheev Jun 18 '12 at 13:30
  • $\begingroup$ For this, I work in RL and interested how people learn the optimal strategy. That is, I havent worked with this task before, but interested in it as it would allow a number of experimental manipulations. I find it hard, and didnt like Dehaene & Changeux (1991), because it doesnt seem to capture the role of memory in the optimal strategy. People can keep track of some, but not very much, history of trials when doing the task. I.e the mathematically optimal strategy seems too computationally heavy, and I couldnt find anyone who developed a better model. Thank you so much for your response $\endgroup$ – user865 Jun 18 '12 at 16:14
  • 3
    $\begingroup$ next time you should include such details in your question (although editing it in now would be a pretty big change to the question). However, take a close look at Kaplan et al. (2006). Hopfield nets are pretty bad at remembering things, and so this approach might be what you are looking for. $\endgroup$ – Artem Kaznatcheev Jun 18 '12 at 18:50
  • $\begingroup$ edited question to incorporate comments $\endgroup$ – Jeromy Anglim Jun 18 '12 at 23:59
  • 2
    $\begingroup$ @ArtemKaznatcheev Sorry about that. I was trying to make the question as useful as possible for future readers. I edited it in line with the exchange in comments that you had with OP. I don't think future readers should have to scour through the comments to understand the question. I think in general where someone can be bothered, insight gained through comments should generally be edited into the question. The question was also looking like it was going to be closed. I have a preference to salvage. I've tweaked the edit a little bit to try to show the changes more clearly. $\endgroup$ – Jeromy Anglim Jun 19 '12 at 0:13

Dehaene & Changeux (1991) made a neural-network model:

The coding units are clusters of neurons organized in layers, or assemblies. A sensonmotor loop enables the network to sort the input cards according to several criteria (color, form, etc.). A higher-level assembly of rule-coding clusters codes for the currently tested rule, which shifts when negative reward is received. Internal testing of the possible rules, analogous to a reasoning process, also occurs, by means of an endogenous autoevaluation loop. When lesioned, the model reproduces the behavior of frontal lobe patients.

Parks et al. (1992) extended the previous neural models of WCST to take into account verbal-fluency.

Amos (2000) built a neural-network model that helps distinguish between the sort of errors made by patients with schizophrenia, Parkinson's disease, and Huntington's disease. He relates the model to neuroanatomy and says:

The model also made specific, empirically falsifiable predictions that can be used to explore the utility of these putative mechanisms of information processing in the frontal cortex and basal ganglia.

Monchi et al. (2000) followed a similar approach as Amos. Their model suggested different impairments in Parkinson’s disease and schizophrenic patients, and made specific predictions of what would be observed in fMRI scans. They tested this prediction in their paper.

By 2005, the WCST had become a benchmark for more general models. Rougier et al. (2005) made a general model of the prefrontal cortex based on general neurobiological principles as opposed to symbolic approaches. They model learnt from experience and generalized to novel tasks. They tested it on WCST and the Stroop task data for typical and frontally damaged subjects.

Kaplan et al. (2006) use a neural-network approach with two parts: a Hopfield net serves as the working memory and a Hamming block as a hypothesis generator.

Bishara et al. (2010) use a more ACT-R-like (although not actual ACT-R) model with probabilistic loading on symbolic rules. The resulting sequential learning model is used to identify specific process that subjects might be struggling with. It is suppose to help diagnosis in a clinical setting, and they test it on substance dependent individuals.

Rigotti et al. (2010) created a network of randomly connected neurons that could solve the task. The key point was that the random connections induced mixed selectivity, which would solve the task with high probability.

| improve this answer | |

My colleagues have applied the COVIS model of category learning to the WCST. COVIS isn't a model of WCST performance per se, but can account for several known phenomena. See this Google Scholar search: http://scholar.google.com/scholar?hl=en&q=Helie+Paul+ashby&btnG=&as_sdt=1%2C5&as_sdtp=

| improve this answer | |
  • 5
    $\begingroup$ Welcome to the site. Are there any articles, in particular, you would recommend? Could you elaborate on how it accounts for known phenomena? $\endgroup$ – Jeromy Anglim Jul 5 '12 at 4:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.