Evolutionary games on the lattice: multitype contact process with density-dependent birth rates

Jonas Köppl, Nicolas Lanchier, Max Mercer · 2024

Read Evolutionary games on the lattice: multitype contact process with density-dependent birth rates by Jonas Köppl, Nicolas Lanchier, Max Mercer (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-19957.txt.

Introduction

Jonas Köppl, Nicolas Lanchier, and Max Mercer rebuild evolutionary game theory on a lattice where sites can sit empty. Their 2024 paper replaces the usual voter-style models, in which every site always holds a player. Their process is a multitype contact process instead. Sites are empty, or hold one of two competing strategies. Players die at a steady rate, and births depend on the local crowding.

Payoffs from neighbors feed the birth rate through a payoff matrix. Doing well against the neighborhood literally breeds. Coupling arguments and block constructions carry the analysis. The authors aim for games that breathe, with territory won and lost rather than labels merely flipping.

Catalog record

CREATORJonas Köppl, Nicolas Lanchier, Max Mercer
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-19957

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Evolutionary games on the lattice: multitype contact process with density-dependent birth rates Jonas Köppl, Nicolas Lanchier, Max Mercer Abstract Interacting particle systems of interest in evolutionary game theory introduced in the probability literature consist of variants of the voter model in which each site is occupied by one player. The goal of this paper is to initiate the study of evolutionary games based more realistically on the multitype contact process in which each site is either empty or occupied by a player following one of two possible competing strategies. Like in the symmetric multitype contact process, players have natural death rate one and natural birth rate λ \lambda . Following the traditional modeling approach of evolutionary game theory, the process also depends on a payoff matrix A = ( a i ​ j ) A=(a_{ij}) where a i ​ j a_{ij} represents the payoff a type i i player receives from each of its type j j neighbors, and the actual birth rate is an increasing function of the payoff. Using various couplings and block constructions, we first prove the existence of a phase transition in the direction of the intra payoff a 11 a_{11} or a 22 a_{22} while the other three payoffs are fixed. We also look at the behavior near the critical point where all four payoffs are equal to zero, in which case the system reduces to the symmetric multitype contact process. The effects of the intra payoffs a 11 a_{11} and a 22 a_{22} are studied using various couplings and duality techniques, while the effects of the inter payoffs a 12 a_{12} and a 21 a_{21} are studied in one dimension using a coupling with the contact process to control the interface between the 1s and the 2s. † † footnotetext: AMS 2000 subject classifications: Primary 60K35, 91A22 Keywords and phrases: Multitype contact process; Evolutionary game theory; Block construction; Duality. 1 Introduction The field of evolutionary game theory was developed by Maynard Smith [ 24 ] , and first appeared in his work with Price [ 25 ] . The basic idea in this field is to reinterpret the different strategies as species and their payoff as fitness to create realistic dynamical systems with density-dependent birth and/or death rates. The earliest model in the field of evolutionary game theory is the (nonspatial deterministic) replicator equation. Having an n\times n payoff matrix A=(a_{ij}) where a_{ij} represents the payoff a type i player receives from a type j player, and letting u_{i} denote the density of type i players in the population, the payoff of each type i player is given by \phi_{i}=\phi_{i}(u_{1},u_{2},\ldots,u_{n})=a_{i1}u_{1}+a_{i2}u_{2}+\cdots+a_{in}u_{n}. Reinterpreting the payoff as fitness (a birth rate when the payoff is positive or minus a death rate when the payoff is negative) and assuming that each individual produced replaces a player chosen uniformly at random, while each individual removed is replaced by a player chosen uniformly at random, result in the following so-called replicator equation [ 16 ] : \begin{array}[]{c}u_{i}^{\prime}=(\phi_{i}u_{i})(\sum_{j\neq i}u_{j})-(\sum_{j\neq i}\phi_{j}u_{j})\,u_{i}=\sum_{j\neq i}(\phi_{i}-\phi_{j})\,u_{i}u_{j}\quad\hbox{for}\quad i=1,2,\ldots,n.\end{array} This system of coupled differential equations can be turned into a spatially explicit stochastic process following the modeling approach of [ 28 , 29 ] . More precisely, to include a spatial structure in the form of local interactions, we first assume that the players are located on the d -dimensional integer lattice \mathbb{Z}^{d} , making the state at time t a spatial configuration \xi_{t}:\mathbb{Z}^{d}\longrightarrow\{1,2,\ldots,n\}\quad\hbox{where}\quad\xi_{t}(x)=\hbox{strategy of the player at site\penalty\ $x$}. Then, writing x\sim y to indicate that the two lattice points x and y are nearest neighbors (distance one apart), the payoff of the player at site x is defined as \begin{array}[]{c}\phi(x,\xi_{t})=\sum_{i,j}a_{ij}\,f_{j}(x,\xi_{t})\,\mathbf{1}\{\xi_{t}(x)=i\}\quad\hbox{where}\quad f_{j}(x,\xi_{t})=\sum_{y\sim x}\mathbf{1}\{\xi_{t}(y)=j\}/2d\end{array} (1) denotes the fraction of nearest neighbors of site x following strategy j , i.e., the payoff only depends on the strategy of the neighbors. The fitness of the player at x is then given by \Phi(x,\xi_{t})=(1-w)\times 1+w\times\phi(x,\xi_{t})=(1-w)\times 1+w\times\hbox{payoff}, where the parameter w\in[0,1] represents the strength of selection. Weak selection refers to the case where w is small, while strong selection means that w=1 . The most popular models that fall in this framework are the birth-death updating process and the death-birth updating process introduced in [ 30 ] . In the birth-death updating process, the fitness is interpreted as a birth rate, and offspring replace a neighbor of the parent’s site chosen uniformly at random, so the rate at which site x switches from strategy i to strategy j is given by \begin{array}[]{c}c_{i\to j}(x,\xi_{t})=\sum_{y\sim x}\Phi(y,\xi_{t})\,\mathbf{1}\{\xi_{t}(y)=j\}/2d\quad\hbox{for all}\quad i\neq j.\end{array} (2) In contrast, in the death-birth updating process, players die at rate one and are instantaneously replaced by the offspring of a neighbor chosen at random with a probability proportional to its fitness, so the local transition rates are given by \begin{array}[]{c}c_{i\to j}(x,\xi_{t})=\sum_{y\sim x}\Phi(y,\xi_{t})\,\mathbf{1}\{\xi_{t}(y)=j\}/\sum_{y\sim x}\Phi(y,\xi_{t})\quad\hbox{for all}\quad i\neq j.\end{array} (3) Taking w=0 , the transition rates in ( 2 )–( 3 ) simplify to f_{j}(x,\xi_{t}) , showing that both processes reduce to the voter model [ 4 , 17 ] . In the presence of weak selection, these two processes were studied in [ 2 , 3 , 23 ] in the context of two-strategy games, while [ 5 , 9 , 26 ] also considered games with more strategies such as rock-paper-scissors. In the weak selection limit w\to 0 , voter model perturbations techniques developed in [ 6 ] can be used to have a precise description of the phase structure of the processes. In the presence of strong selection w=1 , more qualitative aspects such as the existence of phase transitions were proved in [ 12 , 19 ] . Other natural variants of these models with discontinuous transition rates were also studied rigorously in the presence of strong selection: the best-response dynamics [ 11 ] , and the death-birth of the fittest process [ 13 ] . For more details about these models, we refer the reader to [ 21 , Chapter 7] . 2 Model description and main results In the previous models, each birth/death induces the instantaneous death/birth of a neighbor to ensure that each site is occupied by exactly one player at all times. These models and all the models of interacting particle systems of interest in evolutionary game theory that have been studied in the probability literature, consist of variants of the voter model with density-dependent birth and/or death rates. As far as we know, the only exception is the variant of Neuhauser’s multitype contact process [ 27 ] introduced in [ 20 ] . This process, however, was only designed to model the interactions among cooperators and defectors in the prisoner’s dilemma rather than general games described by a payoff matrix. The main objective of this paper is to initiate the study of spatial evolutionary games based more realistically on the multitype contact process instead of the voter model. In particular, focusing for simplicity on two-strategy games, the state at time t is now \xi_{t}:\mathbb{Z}^{d}\longrightarrow\{0,1,2\}\quad\hbox{where}\quad\xi_{t}(x)=\hbox{strategy of the player at site\penalty\ $x$}, with the convention 0 = empty. The dynamics combines the dynamics of the multitype contact process and the dynamics of the birth-death updating process ( 2 ). Like in the symmetric multitype contact process, we assume that, regardless of their strategy, the players have the same natural

Read the complete text