Source-linked AI summary
Disentangling high-order mechanisms and high-order behaviours in complex systems
Fernando E. Rosas, Pedro A. M. Mediano, Andrea I. Luppi, Thomas F. Varley, Joseph T. Lizier, Sebastiano Stramaglia, Henrik J. Jensen, Daniele Marinazzo
TL;DR
Research on high-order mechanisms and high-order behaviours addresses distinct, non-interchangeable questions about complex systems. This correspondence clarifies their complementarity and argues that distinguishing them helps guide hypotheses and tool choice.
Problem
High-order mechanisms and behaviours are distinct questions, but treating behaviour as necessarily based on high-order mechanisms is mistaken; pairwise-only statistics may miss information present in joint distributions.
Source: p. 1Method
The paper distinguishes mechanisms in the data-generating process from emergent behaviours in multivariate statistics, drawing on information-theoretic work on high-order interdependencies.
Source: p. 1Results
The paper concludes that high-order mechanisms and behaviours are complementary perspectives rather than competing approaches to high-order phenomena.
Source: p. 1Takeaways & Limitations
Distinguishing the two perspectives supports more precise hypotheses and tool choices for probing high-order phenomena.
Source: p. 1The distinction is presented as a dominant trend, while some techniques can represent both mechanisms and behaviours.
Source: p. 1
Abstract
from arXiv · showhide
Battiston et al. (arXiv:2110.06023) provide a comprehensive overview of how investigations of complex systems should take into account interactions between more than two elements, which can be modelled by hypergraphs and studied via topological data analysis. Following a separate line of enquiry, a broad literature has developed information-theoretic tools to characterize high-order interdependencies from observed data. While these could seem to be competing approaches aiming to address the same question, in this correspondence we clarify that this is not the case, and that a complete account of higher-order phenomena needs to embrace both.