On the importance of feedback for categorization: Revisiting category learning experiments using an adaptive filter...

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Machine Learning

Background:

• Category learning research has largely shifted from associative to cognitive explanations.
• Cognitive models emphasize similarity and explicit rules over associative mechanisms.
• Previous associative models faced challenges explaining certain category learning phenomena.

Purpose of the Study:

• To implement and test an Adaptive Linear Filter (ALF) model.
• To evaluate ALF's ability to account for category learning tasks problematic for associative views.
• To demonstrate the generality of the ALF model across diverse learning scenarios.

Main Methods:

• Computational simulations using an Adaptive Linear Filter (ALF).
• ALF implementation is closely related to the Rescorla and Wagner learning rule.
• Testing the ALF model on three distinct category learning tasks.

Main Results:

• The ALF model successfully predicted outcomes on challenging category learning tasks.
• Results indicate ALF can account for phenomena previously attributed to cognitive processes.
• Consistent performance across simulations highlights the model's generality.

Conclusions:

• Associative learning principles, as implemented in ALF, remain viable for explaining complex category learning.
• The findings challenge the complete abandonment of associative accounts in favor of cognitive explanations.
• This work has significant implications for the broader category learning literature.