Cellular cooperation with shift updating and repulsion

Andreas Pavlogiannis, Krishnendu Chatterjee, Ben Adlam, Martin A. NowakScientific 2015 · DOI

Abstract

Population structure can facilitate evolution of cooperation. In a structured population, cooperators can form clusters which resist exploitation by defectors. Recently, it was observed that a shift update rule is an extremely strong amplifier of cooperation in a one dimensional spatial model. For the shift update rule, an individual is chosen for reproduction proportional to fecundity; the offspring is placed next to the parent; a random individual dies. Subsequently, the population is rearranged (shifted) until all individual cells are again evenly spaced out. For large population size and a one dimensional population structure, the shift update rule favors cooperation for any benefit-to-cost ratio greater than one. But every attempt to generalize shift updating to higher dimensions while maintaining its strong effect has failed. The reason is that in two dimensions the clusters are fragmented by the movements caused by rearranging the cells. Here we introduce the natural phenomenon of a repulsive force between cells of different types. After a birth and death event, the cells are being rearranged minimizing the overall energy expenditure. If the repulsive force is sufficiently high, shift becomes a strong promoter of cooperation in two dimensions.

Introduction: The Fragmentation Problem in 2D

In evolutionary game theory, the "Prisoner’s Dilemma" serves as the foundational model for studying the emergence of cooperation. Natural selection typically favors "defectors"—those who reap benefits without incurring costs—unless a mechanism exists to protect "cooperators." Spatial structure is one such mechanism; on a grid, cooperators can form clusters that shield their members from exploitation.

A specific mechanism called the shift update rule recently emerged as a powerful amplifier of cooperation in 1D cycles. In this model, when a cell reproduces, its offspring is placed adjacent to it, a random cell elsewhere in the population dies, and the remaining cells "shift" to fill the vacancy. In 1D, this favors cooperation for nearly any benefit-to-cost ($b/c$) ratio greater than one. However, transitioning this rule to 2D grids has historically failed. In two dimensions, the shifting process acts as a "blender," fragmenting cooperator clusters and allowing defectors to invade. This research introduces a physical principle—repulsion between different cell types—to solve this fragmentation problem.

Methods: Introducing Repulsion and Path Optimization

The authors propose a Generic Shift Rule applied to a 2D torus (a wrapping grid). The innovation lies in a repulsion force ($\alpha$) between different cell types (Cooperator vs. Defector). When a birth/death event occurs, the path taken to rearrange the population is no longer random but is determined by the minimization of "energy" or resistance.

The shift update process in a 2D grid showing reproduction, death, and the resulting shift path.

Two specific path-finding rules were tested:

  1. Least-resistance rule (LR): A global optimization where the path minimizes the total sum of repulsion forces encountered.
  2. Least neighbor-resistance rule (LNR): A local, stochastic optimization where each step in the shift path is more likely to move toward a neighbor with lower resistance.

The study utilized Monte Carlo simulations on grid sizes ranging from $4 \times 4$ to $25 \times 25$ to calculate fixation probabilities under strong selection and mathematical structure coefficients ($\sigma$) for weak selection.

Key Results: Breaking the $b/c$ Barrier

The introduction of repulsion effectively restores the "shielding" effect of clusters. When the repulsion force $\alpha$ is high, the shift path is forced to navigate around clusters of the opposite type rather than cutting through them, preventing the "blender effect" that previously destroyed cooperation in 2D.

Surprisingly, the LNR rule (local optimization) outperformed the LR rule (global optimization). LNR leads to fewer, larger, and more resilient clusters. Under the LNR rule with high repulsion, the critical $b/c$ ratio required for cooperation to be favored dropped to as low as 1.3 on a $20 \times 20$ grid. This is a significant improvement over the standard threshold of $b/c > 4$ (the number of neighbors) found in traditional 2D models like death-birth updating.

Discussion and Future Directions

The failure of previous 2D shift models was essentially topological. By incorporating a "physical" property like repulsion, the authors forced evolutionary dynamics to respect cluster boundaries. This "physicality" allows cooperation to flourish even when benefits are low.

While successful, the model faces computational challenges; the shortest-path calculations required for every event limit the scale of simulations. Future work may explore these dynamics in 3D environments or non-regular network structures. Furthermore, the authors suggest the model could be adapted to "semi-fluid" biological systems, such as biofilms, where the level of "viscosity" or repulsion can be tuned to match real-world observations.

Summary updated

Abstract

Population structure can facilitate evolution of cooperation. In a structured population, cooperators can form clusters which resist exploitation by defectors. Recently, it was observed that a shift update rule is an extremely strong amplifier of cooperation in a one dimensional spatial model. For the shift update rule, an individual is chosen for reproduction proportional to fecundity; the offspring is placed next to the parent; a random individual dies. Subsequently, the population is rearranged (shifted) until all individual cells are again evenly spaced out. For large population size and a one dimensional population structure, the shift update rule favors cooperation for any benefit-to-cost ratio greater than one. But every attempt to generalize shift updating to higher dimensions while maintaining its strong effect has failed. The reason is that in two dimensions the clusters are fragmented by the movements caused by rearranging the cells. Here we introduce the natural phenomenon of a repulsive force between cells of different types. After a birth and death event, the cells are being rearranged minimizing the overall energy expenditure. If the repulsive force is sufficiently high, shift becomes a strong promoter of cooperation in two dimensions.
BibTeX
@inproceedings{pavlogiannis2015cellular,
  author = {Andreas Pavlogiannis and Krishnendu Chatterjee and Ben Adlam and Martin A. Nowak},
  title = {{Cellular cooperation with shift updating and repulsion}},
  booktitle = {Scientific},
  year = {2015},
  doi = {10.1038/srep17147},
}
Figure