Beyond networks, towards adaptive systems

TL;DR

This paper critiques traditional network models in neuroscience, highlighting their limitations in capturing biological systems' adaptive, dynamic, and context-dependent nature, and advocates for more flexible frameworks inspired by concepts like the adjacent possible.

Contribution

It identifies key limitations of existing network models in biology and proposes the development of new, more adaptable frameworks that better represent complex biological dynamics.

Findings

Traditional networks struggle with biological complexity

Extensions like multilayer networks partially address limitations

Biological systems require models that handle open-endedness and context dependence

Abstract

Despite their widespread utility across domains, basic network models face fundamental limitations when applied to complex biological systems, particularly in neuroscience. This paper critically examines these limitations and explores potential extensions and alternative frameworks better suited to capture the adaptive nature of biological systems. Key challenges include: the need to account for time-varying connections and adaptive topologies; the difficulty in representing multilevel systems with cross-level interactions; the inadequacy of fixed state spaces for systems that continuously expand their range of possible states; and the challenge of modeling deep history dependence and open-world interactions. While some limitations can be partially addressed through extensions like multilayer networks and time-varying connections, I argue that biological systems exhibit radical context dependence and open-endedness that resist conventional mathematical formalization. Using examples primarily from neuroscience, I demonstrate how even sophisticated approaches incorporating dynamics and context sensitivity fall short in fully capturing biological complexity. Moving forward requires developing frameworks that explicitly account for adaptive reconfiguration, embrace open-ended state spaces, and draw inspiration from concepts like Kauffman's "adjacent possible." The path ahead demands mathematical and computational approaches that are fundamentally more dynamic and flexible than traditional network models.