Related Experiment Video
Updated: Aug 26, 2025
05:55
Modeling the Functional Network for Spatial Navigation in the Human Brain
Published on: October 13, 2023
1.2K
Degrees of algorithmic equivalence between the brain and its DNN models
Philippe G Schyns1, Lukas Snoek1, Christoph Daube1
1School of Psychology and Neuroscience, University of Glasgow, Glasgow G12 8QB, UK.
Trends in Cognitive Sciences
|October 10, 2022
Summary
Deep neural networks (DNNs) model human cognition but may not use similar algorithms. This study proposes benchmarks to evaluate algorithmic similarity between DNNs and human categorization.
Area of Science:
- Cognitive Science
- Neuroscience
- Artificial Intelligence
Background:
- Deep neural networks (DNNs) are increasingly used to model human cognition, showing similar behavioral outcomes in tasks like image categorization.
- The hierarchical, brain-inspired structure of DNNs leads to questions about whether their underlying categorization algorithms mirror human processes.
Purpose of the Study:
- To investigate the extent to which deep neural networks (DNNs) share algorithmic similarities with human cognition.
- To develop a framework for evaluating algorithmic equivalence between DNNs and human cognitive processes.
Main Methods:
- Framed the evaluation of algorithmic similarity in three progressively constrained degrees: behavioral equivalence, stimulus feature processing equivalence, and algorithmic processing equivalence.
- Developed a benchmark with specified epistemological conditions for each degree of equivalence.
Main Results:
- Current practice primarily focuses on behavioral/brain response equivalence, which is the least constraining.
- The proposed benchmark offers increasingly stringent criteria to assess deeper levels of algorithmic similarity.
Conclusions:
- Achieving true cognitive modeling with DNNs requires evaluating beyond behavioral outputs to understand shared stimulus features and processing algorithms.
- The developed benchmark provides a rigorous approach to advance DNNs as more accurate models of human cognition.
