The interpretation of behavior-model correlations in unidentified cognitive models

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psychology

Background:

• Computational cognitive models are increasingly used to quantify individual differences.
• Parameter estimation is a common application for these models.
• Model identification is crucial for valid inferences about individual behavior.

Purpose of the Study:

• To examine the implications of the scaling constraint in computational model identification.
• To caution against overinterpretation of results derived from models with scaling constraints.
• To provide guidance for clearer interpretation of computational modeling findings.

Main Methods:

• Analysis of model identification in computational cognitive models.
• Illustrative examples using Signal Detection Theory, Reinforcement Learning models, and the Linear Ballistic Accumulator model.
• Discussion of the assumptions implied by scaling constraints.

Main Results:

• Model identification is often achieved up to a scaling constraint.
• This constraint imposes a strong assumption on the cognitive processes modeled.
• Overinterpretation of associative relations derived from such models is a potential pitfall.

Conclusions:

• The scaling constraint in model identification requires careful consideration.
• Researchers should be cautious about overinterpreting findings based solely on parameter estimates under scaling constraints.
• Clearer interpretation guidelines are needed for computational modeling results.