Information flow in a network model and the law of diminishing marginal\n returns

Daniele Marinazzo, M. Pellicoro, Guorong Wu, Leonardo Angelini, Sebastiano Stramaglia

Abstract

We analyze a simple dynamical network model which describes the limited capacity of nodes to process the input information. For a suitable choice of the parameters, the information flow pattern is characterized by exponential distribution of the incoming information and a fat-tailed distribution of the outgoing information, as a signature of the law of diminishing marginal returns. The analysis of a real EEG data-set shows that similar phenomena may be relevant for brain signals.

Introduction

In complex networks—whether social, technological, or biological—it is often assumed that increased connectivity translates directly to increased influence. However, every node, such as a person or a neuron, possesses a finite capacity to process incoming information. This paper explores how these individual processing constraints affect the global topology of information flow, demonstrating that the economic law of diminishing marginal returns manifests as a detectable pattern in complex systems.

Methodology

The researchers investigated information dynamics using three primary components:

  1. The Threshold Model: A network where node states evolve based on the sum of neighbors' previous states, passed through a transfer function with a threshold ($\theta$). This mimics saturation in processing capacity.
  2. Multivariate Granger Causality: A statistical tool used to quantify directed information flow ($c_{in}$ and $c_{out}$) by measuring how much the past of one variable improves the prediction of another.
  3. Network Topologies: The models were tested on scale-free (preferential attachment) and homogeneous networks to observe how structure interacts with processing limits.
  4. EEG Data: Recordings from 19 channels in 10 healthy subjects were analyzed to validate the model against real biological signals.

The Hub Paradox and Information Asymmetry

A central finding is the "Hub Paradox." In standard scale-free networks, hubs are expected to be primary information sinks and sources. However, due to processing thresholds, hubs "waste" much of their incoming data once the threshold is exceeded. Consequently, the incoming information ($c_{in}$) becomes nearly uniform across the network, regardless of a node's degree.

Information flow $c$ as a function of node degree $k$. While outgoing information $c_{out}$ (squares) increases with connectivity, incoming information $c_{in}$ (asterisks) remains flat, illustrating the processing bottleneck.

While $c_{in}$ is capped and follows a narrow distribution, the outgoing information ($c_{out}$) remains tied to connectivity. Hubs still exert significant influence on the rest of the network, leading to a "fat-tailed" distribution of outgoing flow. This asymmetry is a hallmark of the law of diminishing returns: adding more input to a hub yields progressively smaller increases in useful internal information, even as its output remains high.

Distributions of incoming ($\rho_{in}$) and outgoing ($\rho_{out}$) information. The narrow, exponential-like distribution of $c_{in}$ versus the broader, fat-tailed distribution of $c_{out}$ highlights the fundamental asymmetry in information processing.

Empirical Evidence in the Brain

The most striking result is that this exact asymmetry—uniform input versus skewed output—is present in human EEG data. This suggests that the brain's architecture is naturally tuned to handle processing bottlenecks. In the brain, the ratio of outgoing to incoming information variance is highest in temporal regions, suggesting these areas are specialized for managing high-throughput information under the constraints of diminishing returns.

Conclusion

By bridging network science and economic principles, this study reveals that processing limits at the node level reshape global information flow. The transition from uniform input to skewed output serves as a signature of systems operating under the law of diminishing marginal returns, a phenomenon observed in both theoretical models and the human brain.

Summary updated

Abstract

We analyze a simple dynamical network model which describes the limited\ncapacity of nodes to process the input information. For a suitable choice of\nthe parameters, the information flow pattern is characterized by exponential\ndistribution of the incoming information and a fat-tailed distribution of the\noutgoing information, as a signature of the law of diminishing marginal\nreturns. The analysis of a real EEG data-set shows that similar phenomena may\nbe relevant for brain signals.\n
BibTeX
@misc{marinazzo2012information,
  author = {Daniele Marinazzo and M. Pellicoro and Guorong Wu and Leonardo Angelini and Sebastiano Stramaglia},
  title = {{Information flow in a network model and the law of diminishing marginal\textbackslash\{\}n returns}},
  year = {2012},
  eprint = {1202.5041},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1202.5041},
}
Figure