Structural Parameter Interdependencies in Computational Models of Cognition

Computational modeling of cognition allows latent psychological variables to be measured by means of adjustable model parameters. The estimation and interpretation of the parameters are impaired, however, if parameters are strongly intercorrelated within the model. We point out that strong parameter interdependencies are especially likely to emerge in models that combine a subjective value function with a probabilistic choice rule-a common structure in the literature. We trace structural parameter interdependencies between value function and choice rule parameters across several prominent computational models, including models on risky choice (cumulative prospect theory), categorization (the generalized context model), and memory (the SIMPLE model of free recall). Using simulation studies with a generic choice model, we show that the accuracy in parameter estimation is hampered in the presence of high parameter intercorrelations, particularly the ability to detect group differences on the parameters and associations of the parameters with external variables. We demonstrate that these problems can be alleviated by using a different specification of stochasticity in the model, for example, by assuming parameter stochasticity or a constant error term. In addition, application to two empirical data sets of risky choice shows that alleviating parameter interdependencies in this way can lead to different conclusions about the estimated parameters. Our analyses highlight a common but often neglected problem of computational models of cognition and identify ways in which the design and application of such models can be improved.