Random Choices can Facilitate the Solving of Collective Network Coloring Problems by Artificial Agents

TL;DR

This paper investigates how introducing randomness in decision-making of artificial agents can improve their ability to solve network coloring problems, emphasizing the importance of tuning randomness based on network structure.

Contribution

It provides analytical and simulation-based evidence that behavioral randomness enhances collective problem-solving in network coloring, with insights into optimal randomness implementation.

Findings

Randomness improves conflict resolution in network coloring.

Optimal noise levels depend on network size and structure.

Behavioral randomness helps avoid local minima in distributed algorithms.

Abstract

Global coordination is required to solve a wide variety of challenging collective action problems from network colorings to the tragedy of the commons. Recent empirical study shows that the presence of a few noisy autonomous agents can greatly improve collective performance of humans in solving networked color coordination games. To provide further analytical insights into the role of behavioral randomness, here we study myopic artificial agents attempt to solve similar network coloring problems using decision update rules that are only based on local information but allow random choices at various stages of their heuristic reasonings. We consider that agents are distributed over a random bipartite network which is guaranteed to be solvable with two colors. Using agent-based simulations and theoretical analysis, we show that the resulting efficacy of resolving color conflicts is dependent on the specific implementation of random behavior of agents, including the fraction of noisy agents and at which decision stage noise is introduced. Moreover, behavioral randomness can be finely tuned to the specific underlying population structure such as network size and average network degree in order to produce advantageous results in finding collective coloring solutions. Our work demonstrates that distributed greedy optimization algorithms exploiting local information should be deployed in combination with occasional exploration via random choices in order to overcome local minima and achieve global coordination.