Exploring Major Transitions in the Evolution of Biological Cognition With Artificial Neural Networks

2025-09-17

AI Summary

It remains unclear whether cognitive evolution depends on qualitative shifts in information flow induced by major topological reconfigurations of neural architectures. Method: We systematically compared feedforward, recurrent, and hierarchical artificial neural networks—matched for scale and computational resources—on an artificial grammar learning task spanning varying syntactic complexities. Contribution/Results: Only recurrent networks exhibited a qualitative performance transition—abrupt emergence of stable generalization—from failure to success under high-complexity grammars. This transition coincided with sharply increased training difficulty and stochastic, irreversible convergence dynamics—paralleling evolutionary “critical transitions” or fitness valleys. Hierarchical connectivity yielded no analogous effect. Our findings constitute the first controlled computational demonstration that topology-specific information-flow reconfiguration can independently drive qualitative expansions in cognitive capacity. This provides a testable mechanistic account of cognitive evolutionary leaps, linking network architecture to emergent computational capabilities without requiring changes in scale or learning algorithm.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Evolutionary Learning
• Search and Optimization: Evolutionary Computation

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Web Mining and Content Analysis: Models for Web evolution
• Economics, Online Markets and Human Computation: Social networks and social learning

Abstract

Transitional accounts of evolution emphasise a few changes that shape what is evolvable, with dramatic consequences for derived lineages. More recently it has been proposed that cognition might also have evolved via a series of major transitions that manipulate the structure of biological neural networks, fundamentally changing the flow of information. We used idealised models of information flow, artificial neural networks (ANNs), to evaluate whether changes in information flow in a network can yield a transitional change in cognitive performance. We compared networks with feed-forward, recurrent and laminated topologies, and tested their performance learning artificial grammars that differed in complexity, controlling for network size and resources. We documented a qualitative expansion in the types of input that recurrent networks can process compared to feed-forward networks, and a related qualitative increase in performance for learning the most complex grammars. We also noted how the difficulty in training recurrent networks poses a form of transition barrier and contingent irreversibility -- other key features of evolutionary transitions. Not all changes in network topology confer a performance advantage in this task set. Laminated networks did not outperform non-laminated networks in grammar learning. Overall, our findings show how some changes in information flow can yield transitions in cognitive performance.

Problem

• Evaluating network topology changes on cognitive performance
• Comparing feed-forward, recurrent, laminated neural networks
• Testing artificial grammar learning across network architectures

Innovation

• Used artificial neural networks to model cognitive evolution
• Compared feed-forward, recurrent, and laminated network topologies
• Found recurrent networks enable qualitative performance improvements