Mutualisms: cooperation between species - Revision history

Hauert: /* Publications */

2025-03-07T21:08:02Z

Publications

← Older revision		Revision as of 14:08, 7 March 2025
Line 211:		Line 211:

	==Publications==		==Publications==
	# Hauert, C. & Szabó, G. (2024) Spontaneous symmetry breaking of cooperation between species, ''PNAS Nexus'' (~~in print~~) doi: [https://doi.org/10.1093/pnasnexus/pgae326 10.1093/pnasnexus/pgae326]		# Hauert, C. & Szabó, G. (2024) Spontaneous symmetry breaking of cooperation between species, ''PNAS Nexus'' '''3''' (9) pgae326 doi: [https://doi.org/10.1093/pnasnexus/pgae326 10.1093/pnasnexus/pgae326]

	===References===		===References===

Hauert at 09:29, 21 February 2025

2025-02-21T09:29:55Z

← Older revision		Revision as of 02:29, 21 February 2025
Line 10:		Line 10:

	\begin{array}{cc}		\begin{array}{cc}
			\label{eq:dg}
	\tag{1}		\tag{1}
	~~\label{eq:dg}~~
	& C\qquad D\\		& C\qquad D\\
	\begin{matrix}		\begin{matrix}
Line 23:		Line 23:

	Thus, both players prefer mutual cooperation over mutual defection, yet defection yields the higher payoff regardless of what the partner does and hence the conflict of interest arises.		Thus, both players prefer mutual cooperation over mutual defection, yet defection yields the higher payoff regardless of what the partner does and hence the conflict of interest arises.

			Rescaling reduces the interaction in Eq. \eqref{eq:dg} to a single parameter:

			\begin{equation}
			\label{eq:dgr}
			\tag{2}
			{\bf A}=
			\begin{pmatrix} 1-r & -r\\
			1 & 0 \end{pmatrix},
			\end{equation}

			where $r=c/b$ denotes the cost-to-benefit ratio with $0\lt r\lt 1$. Moreover, for dynamics that only depend on payoff differences, a constant can be added to ensure that all payoffs are non-negative:

			\begin{equation}
			\label{eq:dgr+}
			\tag{3}
			{\bf A}=
			\begin{pmatrix} 1 & 0\\
			1+r & r \end{pmatrix}.
			\end{equation}

	The problem of cooperation is exacerbated for actions that bestow benefits not just to other individuals but to those of another species. In particular it is crucial to clearly distinguish between the acts of cooperation of an individual and mutualistic interactions between individuals. Accomplishing mutually beneficial interactions requires coordination of cooperative acts between species.		The problem of cooperation is exacerbated for actions that bestow benefits not just to other individuals but to those of another species. In particular it is crucial to clearly distinguish between the acts of cooperation of an individual and mutualistic interactions between individuals. Accomplishing mutually beneficial interactions requires coordination of cooperative acts between species.
Line 34:		Line 54:
	\begin{align}		\begin{align}
	\label{eq:re}		\label{eq:re}
	\tag{2}		\tag{4}
	\dot x &= x(\xi_C - \bar\xi) = x(1-x)(\xi_C - \xi_D)\\		\dot x &= x(\xi_C - \bar\xi) = x(1-x)(\xi_C - \xi_D)\\
	\end{align}		\end{align}
Line 44:		Line 64:
	\begin{align}		\begin{align}
	\label{eq:remut}		\label{eq:remut}
	\tag{3}		\tag{5}
	\dot x &= x(\xi_C - \bar\xi)\\		\dot x &= x(\xi_C - \bar\xi)\\
			\notag
	\dot y &= y(\zeta_C - \bar\zeta),		\dot y &= y(\zeta_C - \bar\zeta),
	\end{align}		\end{align}
Line 53:		Line 74:
	Eq. \eqref{eq:remut} reduces to $\dot x = x(1-x)(-c_x) \lt 0$ and $\dot y = y(1-y)(-c_y) \lt 0$ such that in both species cooperators dwindle and eventually go extinct. Interestingly, this happens regardless of the magnitude of foregone benefits, $b_x,b_y$. Thus, unsurprisingly the demise of cooperation in inter-species interactions remains unchanged.		Eq. \eqref{eq:remut} reduces to $\dot x = x(1-x)(-c_x) \lt 0$ and $\dot y = y(1-y)(-c_y) \lt 0$ such that in both species cooperators dwindle and eventually go extinct. Interestingly, this happens regardless of the magnitude of foregone benefits, $b_x,b_y$. Thus, unsurprisingly the demise of cooperation in inter-species interactions remains unchanged.

	For simplicity we assume that species $X$ and $Y$ face the same donation game, with costs $c=c_x=c_y$ and benefits $b=b_x=b_y$. Thus, $X$ and $Y$ are interchangeable labels~~. Rescaling reduces the interaction in Eq. \eqref{eq:dg} to a single parameter:~~		For simplicity we assume that species $X$ and $Y$ face the same donation game, with costs $c=c_x=c_y$ and benefits $b=b_x=b_y$. Thus, $X$ and $Y$ are interchangeable labels.

	~~\begin{equation}~~
	~~\label{eq:dgr}~~
	~~\tag{4}~~
	~~{\bf A}=~~
	~~\begin{pmatrix} 1-r & -r\\~~
	~~1 & 0 \end{pmatrix},~~
	~~\end{equation}~~

	~~where $r=c/b$ denotes the cost-to-benefit ratio with $0\lt r\lt 1$~~.

	== Spatial populations ==		== Spatial populations ==

Hauert: /* Conclusions & Discussion */

2024-08-21T15:59:23Z

Conclusions & Discussion

← Older revision		Revision as of 08:59, 21 August 2024
Line 123:		Line 123:

	# for $r \to r_1$ from below, cooperation goes extinct in both layers simultaneously following the directed percolation universality class. Surprisingly, the extinction threshold $r_1\approx0.02283(3)$ is very similar to the threshold in single species interactions $r\approx0.0261$ (estimated from [[#References\|Szabó & Hauert, 2002]]). Even though conditions for cooperation would seem much more challenging in interactions across species.		# for $r \to r_1$ from below, cooperation goes extinct in both layers simultaneously following the directed percolation universality class. Surprisingly, the extinction threshold $r_1\approx0.02283(3)$ is very similar to the threshold in single species interactions $r\approx0.0261$ (estimated from [[#References\|Szabó & Hauert, 2002]]). Even though conditions for cooperation would seem much more challenging in interactions across species.

	# spontaneous symmetry-breaking arises in the frequency of cooperation on the two layers when $r \to r_2$. In fact, two equivalent asymmetric phases exist where individuals in one layer prey on those in the other layer or vice versa. This phase transition is similar to the spontaneous symmetry breaking in the two-dimensional Ising model in the absence of an external magnetic field. Although, the additional degrees of freedom (four instead of two microscopic states) result in deviations in the fluctuations of cooperation frequencies.		# spontaneous symmetry-breaking arises in the frequency of cooperation on the two layers when $r \to r_2$. In fact, two equivalent asymmetric phases exist where individuals in one layer prey on those in the other layer or vice versa. This phase transition is similar to the spontaneous symmetry breaking in the two-dimensional Ising model in the absence of an external magnetic field. Although, the additional degrees of freedom (four instead of two microscopic states) result in deviations in the fluctuations of cooperation frequencies.

	# The third critical transition occurs for $r \to r_3$ from above and is driven by bursts of $DD$ domains. For small $r$ the magnitude of the symmetry breaking decreases with $r$, while the magnitude of bursts increases and causes fluctuations to diverge. However, the identification of the universal features of bursts and the slow transition towards absorbing states requires further, time-consuming simulations and analytical investigations.		# The third critical transition occurs for $r \to r_3$ from above and is driven by bursts of $DD$ domains. For small $r$ the magnitude of the symmetry breaking decreases with $r$, while the magnitude of bursts increases and causes fluctuations to diverge. However, the identification of the universal features of bursts and the slow transition towards absorbing states requires further, time-consuming simulations and analytical investigations.

Hauert: /* Critical phase transitions */

2024-08-11T16:18:49Z

Critical phase transitions

Hauert: /* Conclusions & Discussion */

2024-08-11T06:35:50Z

Conclusions & Discussion

← Older revision		Revision as of 23:35, 10 August 2024
Line 132:		Line 132:
	Symmetry breaking in the frequencies of cooperation between species has been reported before (see [[#References\|Ezoe & Ikegawa, 2013]]). However, the characteristics of the species association is not merely determined by the strategy frequencies but rather by the pairings of individual strategies across layers: mutualistic ($CC$ pairings), exploitative/parasitic ($CD$, $DC$ pairings) or non-existent ($DD$ pairings), i.e. no exchange of goods or services. In social dilemmas individuals fail to achieve the socially optimal outcome out of self interest. Cooperation in mutualistic interactions between species takes more varied forms than in interactions within species. Intriguingly, even in fully symmetrical mutualistic interactions, natural selection may favour asymmetric states where one species exploits the other. The resulting unfair cost allocation is reminiscent of the continuous snowdrift game in intra-species interactions, where evolution can promote the emergence of distinct cooperator and defector traits and spatial settings further promote the diversification into distinct traits (see [[#References\|Hauert & Doebeli, 2021]]).		Symmetry breaking in the frequencies of cooperation between species has been reported before (see [[#References\|Ezoe & Ikegawa, 2013]]). However, the characteristics of the species association is not merely determined by the strategy frequencies but rather by the pairings of individual strategies across layers: mutualistic ($CC$ pairings), exploitative/parasitic ($CD$, $DC$ pairings) or non-existent ($DD$ pairings), i.e. no exchange of goods or services. In social dilemmas individuals fail to achieve the socially optimal outcome out of self interest. Cooperation in mutualistic interactions between species takes more varied forms than in interactions within species. Intriguingly, even in fully symmetrical mutualistic interactions, natural selection may favour asymmetric states where one species exploits the other. The resulting unfair cost allocation is reminiscent of the continuous snowdrift game in intra-species interactions, where evolution can promote the emergence of distinct cooperator and defector traits and spatial settings further promote the diversification into distinct traits (see [[#References\|Hauert & Doebeli, 2021]]).

	In the asymmetric phase, one species produces a social good at much higher rates than the other, which effectively separates the two species into producers and consumers. Interestingly, however, the average level of cooperation across both layers remains essentially constant at $50\%$. Both species produce social goods but the consumers produce theirs at much lower rates. Nevertheless, exploitation in this dynamical equilibrium is limited both spatially and temporally. In the long run, every individual produces one good and receives the other. Surprisingly, this asymmetry tends to become more pronounced for higher stakes, that is for '''decreasing''' cost-to-benefit ratios. Although the social dilemma becomes more benign and defection less tempting, the inequalities between species culminate in full exploitation under the most benign conditions.		In the asymmetric phase, one species produces a social good at much higher rates than the other, which effectively separates the two species into producers and consumers. Interestingly, however, the average level of cooperation across both layers remains essentially constant at $50\%$. Both species produce social goods but the consumers produce theirs at much lower rates. Nevertheless, exploitation in this dynamical equilibrium is limited both spatially and temporally. In the long run, every individual produces one good and receives the other. Surprisingly, this asymmetry tends to become more pronounced for higher stakes, that is for ''decreasing'' cost-to-benefit ratios. Although the social dilemma becomes more benign and defection less tempting, the inequalities between species culminate in full exploitation under the most benign conditions.
	{{-}}		{{-}}

Hauert at 06:35, 11 August 2024

2024-08-11T06:35:07Z

← Older revision		Revision as of 23:35, 10 August 2024
Line 6:		Line 6:
	The evolution and maintenance of cooperation ranks among the most fundamental questions in biological, social and economical systems. Cooperators provide benefits to others at a cost to themselves. Cooperation is a conundrum because on the one hand everyone benefits from mutual cooperation but on the other hand individuals face the temptation to increase their personal gains by defecting and withholding cooperation. This generates a conflict of interest between the group and the individual, which is termed a social dilemma. The prisoner's dilemma is the most popular mathematical metaphor of a social dilemma and theoretical tool to study cooperation through evolutionary game theory.		The evolution and maintenance of cooperation ranks among the most fundamental questions in biological, social and economical systems. Cooperators provide benefits to others at a cost to themselves. Cooperation is a conundrum because on the one hand everyone benefits from mutual cooperation but on the other hand individuals face the temptation to increase their personal gains by defecting and withholding cooperation. This generates a conflict of interest between the group and the individual, which is termed a social dilemma. The prisoner's dilemma is the most popular mathematical metaphor of a social dilemma and theoretical tool to study cooperation through evolutionary game theory.

	== Donation game ==		=== Donation game ===
	In the simplest and most intuitive instance of the prisoner's dilemma, the donation game, two players meet and decide whether to cooperate and provide a benefit $b$ to their interaction partner at a cost $c$ ($b>c>0$) or to defect, which entails no costs and provides no benefits. If both cooperate each gets $b-c$ but if one defects then the defector gets $b$ and shirks the costs of cooperation, while the cooperator is stuck with the costs $-c$. Finally if both defect, no one gets anything. The interaction is conveniently summarized in a payoff matrix (for the row player):		In the simplest and most intuitive instance of the prisoner's dilemma, the donation game, two players meet and decide whether to cooperate and provide a benefit $b$ to their interaction partner at a cost $c$ ($b>c>0$) or to defect, which entails no costs and provides no benefits. If both cooperate each gets $b-c$ but if one defects then the defector gets $b$ and shirks the costs of cooperation, while the cooperator is stuck with the costs $-c$. Finally if both defect, no one gets anything. The interaction is conveniently summarized in a payoff matrix (for the row player):

Line 115:		Line 115:
	However, the $CC$ domains are not stable against invasion of $CD$ or $DC$ phases ('''d''').		However, the $CC$ domains are not stable against invasion of $CD$ or $DC$ phases ('''d''').
	Subsequently, the strategy distribution evolves slowly toward the initial distribution ('''a''').		Subsequently, the strategy distribution evolves slowly toward the initial distribution ('''a''').
			{{-}}

			== Conclusions & Discussion ==
			Interactions between species play a crucial role in symbiotic associations. More specifically, in mutualistic exchanges members of one species provide benefits in the form of goods or services to members of another species and vice versa. For costly provisions this results in an inter-species donation game but with a more fragile arrangement than for the standard, intra-species donation game. Here we studied the evolution of costly acts of cooperation on a two-layer square lattice, where each layer represents one species. This models a basic two-species ecological system. Evolution is controlled by increased reproduction or imitation of fitter neighbours within the same layer, while the payoff (or fitness) of individuals is determined through interactions between layers.

			Using numerical simulations, we determined the frequency and fluctuation of cooperation in both layers as a function of the cost-to-benefit ratio, $r$. This reveals three distinct critical phase transitions:

			# for $r \to r_1$ from below, cooperation goes extinct in both layers simultaneously following the directed percolation universality class. Surprisingly, the extinction threshold $r_1\approx0.02283(3)$ is very similar to the threshold in single species interactions $r\approx0.0261$ (estimated from [[#References\|Szabó & Hauert, 2002]]). Even though conditions for cooperation would seem much more challenging in interactions across species.

			# spontaneous symmetry-breaking arises in the frequency of cooperation on the two layers when $r \to r_2$. In fact, two equivalent asymmetric phases exist where individuals in one layer prey on those in the other layer or vice versa. This phase transition is similar to the spontaneous symmetry breaking in the two-dimensional Ising model in the absence of an external magnetic field. Although, the additional degrees of freedom (four instead of two microscopic states) result in deviations in the fluctuations of cooperation frequencies.

			# The third critical transition occurs for $r \to r_3$ from above and is driven by bursts of $DD$ domains. For small $r$ the magnitude of the symmetry breaking decreases with $r$, while the magnitude of bursts increases and causes fluctuations to diverge. However, the identification of the universal features of bursts and the slow transition towards absorbing states requires further, time-consuming simulations and analytical investigations.

			From a dynamical systems perspective this last critical transition is the most intriguing feature of our model. It poses a challenge for understanding the underlying mechanisms that characterize this universality class and drive the observed power law behaviour. Conversely, from a more applied, biological or social perspective, the spontaneous symmetry breaking and the dynamics of the asymmetric phase are equally clearly the most relevant and compelling feature. In particular, the mechanism for generating and maintaining the asymmetric phase appears to be intrinsically linked to the spikes in defection observed for $r\to r_3$. More specifically, the emergence of asymmetries is driven by cyclical invasions of locally homogeneous domains: $CD$, $DC$ invade $CC$ invades $DD$ invades $CD$, $DC$ etc. This cycle becomes increasingly pronounced for decreasing $r$ because the local domains grow and eventually give rise to the global spikes in defection that characterize the third critical transition. Assuming more $C$'s in one layer, the invading $CC$ domains are more likely to encounter a $D$ in the other layer, which then amplifies the asymmetry.

			Symmetry breaking in the frequencies of cooperation between species has been reported before (see [[#References\|Ezoe & Ikegawa, 2013]]). However, the characteristics of the species association is not merely determined by the strategy frequencies but rather by the pairings of individual strategies across layers: mutualistic ($CC$ pairings), exploitative/parasitic ($CD$, $DC$ pairings) or non-existent ($DD$ pairings), i.e. no exchange of goods or services. In social dilemmas individuals fail to achieve the socially optimal outcome out of self interest. Cooperation in mutualistic interactions between species takes more varied forms than in interactions within species. Intriguingly, even in fully symmetrical mutualistic interactions, natural selection may favour asymmetric states where one species exploits the other. The resulting unfair cost allocation is reminiscent of the continuous snowdrift game in intra-species interactions, where evolution can promote the emergence of distinct cooperator and defector traits and spatial settings further promote the diversification into distinct traits (see [[#References\|Hauert & Doebeli, 2021]]).

			In the asymmetric phase, one species produces a social good at much higher rates than the other, which effectively separates the two species into producers and consumers. Interestingly, however, the average level of cooperation across both layers remains essentially constant at $50\%$. Both species produce social goods but the consumers produce theirs at much lower rates. Nevertheless, exploitation in this dynamical equilibrium is limited both spatially and temporally. In the long run, every individual produces one good and receives the other. Surprisingly, this asymmetry tends to become more pronounced for higher stakes, that is for '''decreasing''' cost-to-benefit ratios. Although the social dilemma becomes more benign and defection less tempting, the inequalities between species culminate in full exploitation under the most benign conditions.
	{{-}}		{{-}}

Line 127:		Line 145:

	{{-}}		{{-}}


	[[Image:Mutualism - cooperation fluctuation.png\|thumb\|300px\|Fluctuations of the frequencies of cooperation, $\chi_a$, in each layer ($\Box$ and $\times$) for $K=0.1$. This reveals three critical phase transitions: $r_1, r_2$ and $r_3$, which are characterized by their asymptotic power law scaling of diverging fluctuations.]]		[[Image:Mutualism - cooperation fluctuation.png\|thumb\|300px\|Fluctuations of the frequencies of cooperation, $\chi_a$, in each layer ($\Box$ and $\times$) for $K=0.1$. This reveals three critical phase transitions: $r_1, r_2$ and $r_3$, which are characterized by their asymptotic power law scaling of diverging fluctuations.]]
Line 181:		Line 198:
	Another peculiarity of the burst dynamics is the feedback between their creation and extinction and the spatial configuration of the asymmetric phase: for $r$ close to $r_3$, the asymmetry of cooperation, $\Phi = \|\bar\rho_{+1} - \bar\rho_{-1}\|$, decreases with $r$ while the burst sizes increase.		Another peculiarity of the burst dynamics is the feedback between their creation and extinction and the spatial configuration of the asymmetric phase: for $r$ close to $r_3$, the asymmetry of cooperation, $\Phi = \|\bar\rho_{+1} - \bar\rho_{-1}\|$, decreases with $r$ while the burst sizes increase.

	For $r<r_3$ the population tends to get absorbed in the fully asymmetric $CD$ or $DC$ states. Once one layer is homogeneous the dynamics in the other layer is dominated by neutral drift because the difference in payoffs between cooperators and defectors merely amounts to $4r$. Hence, for small $r$, the probability to update the strategy essentially reduces to a coin toss with a slight bias in favour of defectors. The probability to fixate in one or the other absorbing state is thus essentially given by the frequency of strategies in the heterogeneous layer at the time the other layer turned homogeneous. For example, if one layer becomes homogeneous in $C$, then the frequency of $C$’s in the other layer is likely $<1/2$, due to the asymmetry. Thus, the $CD$ absorbing state is more likely than $CC$. Conversely, if one layer becomes homogeneous in $D$, then the frequency of $C$’s in the other layer is likely $>1/2$ and hence the $DC$ absorbing state is more likely than $DD$. Finally, it is more likely that one layer becomes homogeneous in $D$ because isolated $C$’s (or small $C$ clusters) are more easily eliminated than $D$’s. Thus, even though all four absorbing states may be reached, in principle, the probabilities to do so are very different and most likely the populations approach heterogeneous absorbing states, whereas homogeneous cooperation across lattices is a highly unlikely outcome.		For $r<r_3$ the population tends to get absorbed in the fully asymmetric $CD$ or $DC$ states. Once one layer is homogeneous the dynamics in the other layer is dominated by neutral drift because the difference in payoffs between cooperators and defectors merely amounts to $4r$. Hence, for small $r$, the probability to update the strategy essentially reduces to a coin toss with a slight bias in favour of defectors. The probability to fixate in one or the other absorbing state is thus essentially given by the frequency of strategies in the heterogeneous layer at the time the other layer turned homogeneous. For example, if one layer becomes homogeneous in $C$, then the frequency of $C$’s in the other layer is likely $<1/2$, due to the asymmetry. Thus, the $CD$ absorbing state is more likely than $CC$. Conversely, if one layer becomes homogeneous in $D$, then the frequency of $C$’s in the other layer is likely $>1/2$ and hence the $DC$ absorbing state is more likely than $DD$. Finally, it is more likely that one layer becomes homogeneous in $D$ because isolated $C$’s (or small $C$ clusters) are more easily eliminated than $D$’s. Thus, even though all four absorbing states may be reached, in principle, the probabilities to do so are very different and most likely the populations approach heterogeneous absorbing states, whereas homogeneous cooperation across lattices is a highly unlikely outcome.

	{{-}}		{{-}}

	== ~~References~~ ==		==Publications==
	# Hauert, C. & Szabó, G. (2024) Spontaneous symmetry breaking of cooperation between species, ''PNAS Nexus'' (in print) doi: [https://doi.org/10.1093/pnasnexus/pgae326 10.1093/pnasnexus/pgae326]		# Hauert, C. & Szabó, G. (2024) Spontaneous symmetry breaking of cooperation between species, ''PNAS Nexus'' (in print) doi: [https://doi.org/10.1093/pnasnexus/pgae326 10.1093/pnasnexus/pgae326]

			===References===
			# Ezoe H. & Ikegawa Y. (2013) Coexistence of mutualists and non-mutualists in a dual-lattice model ''J. their. Biol.'' '''332''' 1-8.
			# Szabó, G. & Hauert, C. (2002) Phase Transitions and Volunteering in Spatial Public Goods Games, ''Phys. Rev. Lett.'' '''89''' (11) 118101.
			# Hauert, C. & Doebeli, M. (2023) Spatial social dilemmas promote diversity ''Proc. Natl. Acad. Sci. USA'' '''118''' (42) e2105252118 [https://doi.org/10.1073/pnas.2105252118 doi: 10.1073/pnas.2105252118].

Hauert: /* Critical phase transitions */

2024-08-11T05:58:39Z

Critical phase transitions

← Older revision		Revision as of 22:58, 10 August 2024
Line 124:		Line 124:
	Each critical phase transition is characterized by, indeed defined by, diverging properties, such as fluctuations or correlation times and lengths. As a consequence it is numerically challenging to determine the exact transition points because the diverging quantities require both large lattices as well as long averaging times.		Each critical phase transition is characterized by, indeed defined by, diverging properties, such as fluctuations or correlation times and lengths. As a consequence it is numerically challenging to determine the exact transition points because the diverging quantities require both large lattices as well as long averaging times.

	Overall, the frequency of cooperation tends to increase as the conditions of cooperation become more benign, i.e. lower costs or higher benefits, which translates into smaller ratios $r$ (as shown on the right). ~~Initially~~ the frequency of cooperation is the same in both lattices. This is not surprising because of the symmetrical setup. However, when decreasing $r$ ~~and~~ cooperation ~~would seem to become~~ less challenging, a surprising spontaneous break of the symmetry is observed, which results in essentially complementary frequencies of cooperators and defectors in the two lattices.		Overall, the frequency of cooperation tends to increase as the conditions of cooperation become more benign, i.e. lower costs or higher benefits, which translates into smaller ratios $r$ (as shown on the right). For large $r$ the frequency of cooperation is the same in both lattices. This is not surprising because of the symmetrical setup. However, when decreasing $r$ cooperation becomes less challenging and a surprising spontaneous break of the symmetry is observed, which results in essentially complementary frequencies of cooperators and defectors in the two lattices.

	{{-}}		{{-}}
Line 131:		Line 131:
	[[Image:Mutualism - cooperation fluctuation.png\|thumb\|300px\|Fluctuations of the frequencies of cooperation, $\chi_a$, in each layer ($\Box$ and $\times$) for $K=0.1$. This reveals three critical phase transitions: $r_1, r_2$ and $r_3$, which are characterized by their asymptotic power law scaling of diverging fluctuations.]]		[[Image:Mutualism - cooperation fluctuation.png\|thumb\|300px\|Fluctuations of the frequencies of cooperation, $\chi_a$, in each layer ($\Box$ and $\times$) for $K=0.1$. This reveals three critical phase transitions: $r_1, r_2$ and $r_3$, which are characterized by their asymptotic power law scaling of diverging fluctuations.]]

	In order to determine the boundaries ~~of the dynamical domains at~~ $r_1$, $r_2$, and $r_3$, in descending order of $r$, as well as to classify the accompanying phase transitions, we consider the fluctuations, $\chi_a$, in layer $a$:		In order to determine the boundaries $r_1$, $r_2$, and $r_3$ of the dynamical domains in descending order of $r$, as well as to classify the accompanying phase transitions, we consider the fluctuations, $\chi_a$, in the frequency of cooperation in in layer $a$:

	\begin{align}		\begin{align}
	\chi_a &= N \langle\left(\rho_a(t) - \bar\rho_a\right)^2 \rangle ~~\,,~~		\chi_a &= N \langle\left(\rho_a(t) - \bar\rho_a\right)^2 \rangle .
	\label{eq:chiav}		\label{eq:chiav}
	\end{align}		\end{align}

	~~where~~ $\chi_a$ becomes constant in the limit $N\to\infty$ and $\chi_a=0$ holds for homogeneous states. The scaling of $\chi_a$ as $r$ approaches a critical phase transition helps to identify the corresponding universality class. The large fluctuations that accompany and indeed characterize the critical phase transitions, require significantly larger lattices (up to $L=1800$ with $\tau_s>10^7$, $\tau_r=2\cdot10^6$).		Note that $\chi_a$ becomes constant in the limit $N\to\infty$ and $\chi_a=0$ holds for homogeneous states. The scaling of $\chi_a$ as $r$ approaches a critical phase transition helps to identify the corresponding universality class. The large fluctuations that accompany and indeed characterize the critical phase transitions, require significantly larger lattices (up to $L=1800$ with $\tau_s>10^7$, $\tau_r=2\cdot10^6$).

	{{-}}		{{-}}

Hauert: /* Critical phase transitions */

2024-08-10T20:56:06Z

Critical phase transitions

← Older revision		Revision as of 13:56, 10 August 2024
Line 120:		Line 120:
	[[Image:Mutualism - cooperation frequency.png\|thumb\|300px\|Frequency of cooperation $\bar\rho_a$ in each layer ($\Box$ and $\times$, respectively) as a function of the cost-to-benefit ratio $r$ for $K=0.1$. The inset enlarges the region of spontaneous symmetry breaking and the arrow indicates the $r$ for which a burst of defection is illustrated above. Simulations for lattices with linear size $L=800$, after a relaxation time $\tau_r=10^5$ and averages over $\tau_s=10^5$ MC steps and up to $L=1800$, $\tau_r=2\cdot10^6$ and $\tau_s=10^7$ in the vicinity of critical points.]]		[[Image:Mutualism - cooperation frequency.png\|thumb\|300px\|Frequency of cooperation $\bar\rho_a$ in each layer ($\Box$ and $\times$, respectively) as a function of the cost-to-benefit ratio $r$ for $K=0.1$. The inset enlarges the region of spontaneous symmetry breaking and the arrow indicates the $r$ for which a burst of defection is illustrated above. Simulations for lattices with linear size $L=800$, after a relaxation time $\tau_r=10^5$ and averages over $\tau_s=10^5$ MC steps and up to $L=1800$, $\tau_r=2\cdot10^6$ and $\tau_s=10^7$ in the vicinity of critical points.]]

	The social dilemma in \~~ref~~{eq:dgr} is in effect for $r>0$ and $D$ dominates (mutual defection is the sole Nash equilibrium). However, in spatial settings homogeneous defection results only for sufficiently large $r>r_1\approx0.02283(3)$. For smaller $r$, the limited local interactions suffice to protect cooperators. The formation of clusters reduces exploitation and increases interactions with other cooperators to make up for losses against defectors. Spatial assortment enables co-existence. More specifically, four dynamical regimes exist that are separated by critical phase transitions at $r_1$, $r_2$, and $r_3$.		The social dilemma in Eq. \eqref{eq:dgr} is in effect for $r>0$ and $D$ dominates (mutual defection is the sole Nash equilibrium). However, in spatial settings homogeneous defection results only for sufficiently large $r>r_1\approx0.02283(3)$. For smaller $r$, the limited local interactions suffice to protect cooperators. The formation of clusters reduces exploitation and increases interactions with other cooperators to make up for losses against defectors. Spatial assortment enables co-existence. More specifically, four dynamical regimes exist that are separated by critical phase transitions at $r_1$, $r_2$, and $r_3$.

	Each critical phase transition is characterized by, indeed defined by, diverging properties, such as fluctuations or correlation times and lengths. As a consequence it is numerically challenging to determine the exact transition points because the diverging quantities require both large lattices as well as long averaging times.		Each critical phase transition is characterized by, indeed defined by, diverging properties, such as fluctuations or correlation times and lengths. As a consequence it is numerically challenging to determine the exact transition points because the diverging quantities require both large lattices as well as long averaging times.

Hauert: /* Critical phase transitions */

2024-08-10T20:55:05Z

Critical phase transitions

← Older revision		Revision as of 13:55, 10 August 2024
Line 117:		Line 117:
	{{-}}		{{-}}

	=== Critical phase transitions ===		== Critical phase transitions ==
	[[Image:Mutualism - cooperation frequency.png\|thumb\|300px\|Frequency of cooperation $\bar\rho_a$ in each layer ($\Box$ and $\times$, respectively) as a function of the cost-to-benefit ratio $r$ for $K=0.1$. The inset enlarges the region of spontaneous symmetry breaking and the arrow indicates the $r$ for which a burst of defection is illustrated above. Simulations for lattices with linear size $L=800$, after a relaxation time $\tau_r=10^5$ and averages over $\tau_s=10^5$ MC steps and up to $L=1800$, $\tau_r=2\cdot10^6$ and $\tau_s=10^7$ in the vicinity of critical points.]]		[[Image:Mutualism - cooperation frequency.png\|thumb\|300px\|Frequency of cooperation $\bar\rho_a$ in each layer ($\Box$ and $\times$, respectively) as a function of the cost-to-benefit ratio $r$ for $K=0.1$. The inset enlarges the region of spontaneous symmetry breaking and the arrow indicates the $r$ for which a burst of defection is illustrated above. Simulations for lattices with linear size $L=800$, after a relaxation time $\tau_r=10^5$ and averages over $\tau_s=10^5$ MC steps and up to $L=1800$, $\tau_r=2\cdot10^6$ and $\tau_s=10^7$ in the vicinity of critical points.]]

Hauert at 20:54, 10 August 2024

2024-08-10T20:54:35Z

← Older revision		Revision as of 13:54, 10 August 2024
Line 145:		Line 145:
	[[Image:Mutualism - phase transition r1.png\|thumb\|300px\|The scaling of '''a''' $\bar \rho_a$ and '''b''' $\chi_a$ versus $(r_1-r)/r_1$ follows the directed percolation universality class (solid lines indicate theoretical expectations).]]		[[Image:Mutualism - phase transition r1.png\|thumb\|300px\|The scaling of '''a''' $\bar \rho_a$ and '''b''' $\chi_a$ versus $(r_1-r)/r_1$ follows the directed percolation universality class (solid lines indicate theoretical expectations).]]

	Near the extinction threshold, $r_1$, isolated clusters of cooperators perform a branching and annihilating random walk with almost perfect correlations between lattices. The characteristic spatial patterns are captured by snapshots of the configuration in each lattice as well as the pair distributions (see [[#Symmetric cooperation: random walk\| Symmetric cooperation: random walk]]).		Near the extinction threshold, $r_1$, isolated clusters of cooperators perform a branching and annihilating random walk with almost perfect correlations between lattices. The characteristic spatial patterns are captured by snapshots of the configuration in each lattice as well as the pair distributions (see [[#Symmetric cooperation: random walk\|Symmetric cooperation: random walk]]).

	Coordination between lattices is almost perfect with very few $CD$ and $DC$ pairs. This is of the essence for the survival of cooperation. $CC$ interactions are the bulwark against, or at least help to compensate for exploitation by defectors. Interestingly, the emerging spatial configurations are very similar to those in the intra-species donation game on a single lattice.		Coordination between lattices is almost perfect with very few $CD$ and $DC$ pairs. This is of the essence for the survival of cooperation. $CC$ interactions are the bulwark against, or at least help to compensate for exploitation by defectors. Interestingly, the emerging spatial configurations are very similar to those in the intra-species donation game on a single lattice.
Line 151:		Line 151:
	In agreement with theoretical expectations the critical transition at $r_1$ clearly belongs to the two-dimensional directed percolation universality class: in both layers the frequency of cooperation scales with $\bar\rho_a \propto (r_{1}-r)^\beta$ for $\beta=0.580(3)$ and their fluctuations scale with $\chi_a \propto (r_{1}-r)^{-\gamma}$ for $\gamma=0.35(1)$ when $r\to r_1$ from below (as shown on the right)		In agreement with theoretical expectations the critical transition at $r_1$ clearly belongs to the two-dimensional directed percolation universality class: in both layers the frequency of cooperation scales with $\bar\rho_a \propto (r_{1}-r)^\beta$ for $\beta=0.580(3)$ and their fluctuations scale with $\chi_a \propto (r_{1}-r)^{-\gamma}$ for $\gamma=0.35(1)$ when $r\to r_1$ from below (as shown on the right)

	When lowering the cost-to-benefit ratio, $r_2<r<r_1$, the frequency of cooperation gradually increases with decreasing $r$, while the coordination between layers decreases. For $r$ close to $r_2$ the frequency of all four strategy pairs are essentially the same, which supports that configurations are largely uncorrelated.		When lowering the cost-to-benefit ratio, $r_2<r<r_1$, the frequency of cooperation gradually increases with decreasing $r$, while the coordination between layers decreases. For $r$ close to $r_2$ the frequency of all four strategy pairs are essentially the same, which supports that configurations are largely uncorrelated (see [[#Symmetric cooperation: uncorrelated\|Symmetric cooperation: uncorrelated]]).

	{{-}}		{{-}}
Line 158:		Line 158:
	[[Image:Mutualism - phase transition r2.png\|thumb\|300px\|The scaling of the order parameter $\Phi=\vert\bar\rho_{+1}-\bar\rho_{-1}\vert$ as a function of $(r_2-r)/r_2$ is in line with the universality class of spontaneous magnetization in the Ising model. '''b''' The scaling of the fluctuations of $\Phi$ ($\bullet$ and $\circ$) as well as $\chi_a$ ($\Box$ and $\times$) above and below the critical threshold $r_2$. The $\Phi$ data reproduces the theoretical expectations (solid lines with slopes $1/8$ in '''a''' and $-7/4$ in '''b''') within statistical errors (omitted for clarity).]]		[[Image:Mutualism - phase transition r2.png\|thumb\|300px\|The scaling of the order parameter $\Phi=\vert\bar\rho_{+1}-\bar\rho_{-1}\vert$ as a function of $(r_2-r)/r_2$ is in line with the universality class of spontaneous magnetization in the Ising model. '''b''' The scaling of the fluctuations of $\Phi$ ($\bullet$ and $\circ$) as well as $\chi_a$ ($\Box$ and $\times$) above and below the critical threshold $r_2$. The $\Phi$ data reproduces the theoretical expectations (solid lines with slopes $1/8$ in '''a''' and $-7/4$ in '''b''') within statistical errors (omitted for clarity).]]

	Intriguingly, lowering $r$ further to $r_3<r<r_2$ with $r_3\approx0.00041(3)$ results in spontaneous symmetry breaking with different frequencies in each layer.		Intriguingly, lowering $r$ further to $r_3<r<r_2$ with $r_3\approx0.00041(3)$ results in spontaneous symmetry breaking with different frequencies in each layer (see [[#Asymmetric cooperation\|Asymmetric cooperation]]).

	The distribution of cooperators and defectors in each layer is almost complementary. Clusters, or regions of cooperators, in one layer are matched by defectors in the other. As a consequence, the lattices consist mostly of $CD$ and $DC$ pairs. Because of the distinctly different frequencies of cooperation in each lattice, either $CD$ or $DC$ pairs dominate. However, note that which pair dominates is of no consequence, because species interactions are symmetric, and hence it is merely a consequence of which species is labelled $X$ and which $Y$. Defectors in one species manage to exploit and take advantage of cooperators in the other species. However, exploitation is ephemeral and the scale is limited both spatially as well as temporarily.		The distribution of cooperators and defectors in each layer is almost complementary. Clusters, or regions of cooperators, in one layer are matched by defectors in the other. As a consequence, the lattices consist mostly of $CD$ and $DC$ pairs. Because of the distinctly different frequencies of cooperation in each lattice, either $CD$ or $DC$ pairs dominate. However, note that which pair dominates is of no consequence, because species interactions are symmetric, and hence it is merely a consequence of which species is labelled $X$ and which $Y$. Defectors in one species manage to exploit and take advantage of cooperators in the other species. However, exploitation is ephemeral and the scale is limited both spatially as well as temporarily.
Line 172:		Line 172:
	==== Bursts of defection ====		==== Bursts of defection ====
	[[Image:Mutualism - phase transition r3.png\|thumb\|150px\|<div style="width:296px">Scaling of $\chi_a$ in each layer ($\Box$ and $\times$) versus $(r-r_3)/r_3$, which is driven by bursts of defection. For reference the solid line shows a power law with exponent $-1.33$.</div>]]		[[Image:Mutualism - phase transition r3.png\|thumb\|150px\|<div style="width:296px">Scaling of $\chi_a$ in each layer ($\Box$ and $\times$) versus $(r-r_3)/r_3$, which is driven by bursts of defection. For reference the solid line shows a power law with exponent $-1.33$.</div>]]

			Finally, for $0<r<r_3$, the population relaxes into one of its four absorbing states depending on the initial configuration. In fact, the most striking dynamical phenomenon of this model manifests itself in the diverging fluctuations of $C$ frequencies, $\chi_a$, for $r\to r_3$ (see [[#Bursts of defection\|Bursts of defection]]).

			In particular, no theoretical arguments exist to date that explain and justify the observed power law behaviour of $\chi_a \propto (r-r_3)^{-\gamma}$ for $r\to r_3$ from above with $\gamma=1.33(5)$. The divergence of $\chi_a$ is related to the emergence, growth and elimination of homogeneous $DD$ domains. Spikes in $DD$ frequencies drive the dynamics of bursts.

			This type of bursts are driven by sufficiently large spatial domains of an unstable phase. Evidently, their extension and frequency depends on $r$ and other parameters. In particular, bursts become larger and rarer when decreasing $r$. Once a burst reaches sizes comparable to the entire lattice, it can drive the evolution towards absorbing states. Interestingly, the absorbing $DD$ state is rarely observed because it is prone to invasion by any remaining $CC$ islands, which then pave the way for $CD$ or $DC$ phases to grow and take over. This is in stark contrast to intra-species models where cooperation continuously increases for decreasing $r$, that is as the social dilemma becomes more benign.

			Another peculiarity of the burst dynamics is the feedback between their creation and extinction and the spatial configuration of the asymmetric phase: for $r$ close to $r_3$, the asymmetry of cooperation, $\Phi = \|\bar\rho_{+1} - \bar\rho_{-1}\|$, decreases with $r$ while the burst sizes increase.

			For $r<r_3$ the population tends to get absorbed in the fully asymmetric $CD$ or $DC$ states. Once one layer is homogeneous the dynamics in the other layer is dominated by neutral drift because the difference in payoffs between cooperators and defectors merely amounts to $4r$. Hence, for small $r$, the probability to update the strategy essentially reduces to a coin toss with a slight bias in favour of defectors. The probability to fixate in one or the other absorbing state is thus essentially given by the frequency of strategies in the heterogeneous layer at the time the other layer turned homogeneous. For example, if one layer becomes homogeneous in $C$, then the frequency of $C$’s in the other layer is likely $<1/2$, due to the asymmetry. Thus, the $CD$ absorbing state is more likely than $CC$. Conversely, if one layer becomes homogeneous in $D$, then the frequency of $C$’s in the other layer is likely $>1/2$ and hence the $DC$ absorbing state is more likely than $DD$. Finally, it is more likely that one layer becomes homogeneous in $D$ because isolated $C$’s (or small $C$ clusters) are more easily eliminated than $D$’s. Thus, even though all four absorbing states may be reached, in principle, the probabilities to do so are very different and most likely the populations approach heterogeneous absorbing states, whereas homogeneous cooperation across lattices is a highly unlikely outcome.

	{{-}}		{{-}}

← Older revision		Revision as of 08:59, 21 August 2024
Line 123:		Line 123:

	# for \(r \to r_1\) from below, cooperation goes extinct in both layers simultaneously following the directed percolation universality class. Surprisingly, the extinction threshold \(r_1\approx0.02283(3)\) is very similar to the threshold in single species interactions \(r\approx0.0261\) (estimated from [[#References\|Szabó & Hauert, 2002]]). Even though conditions for cooperation would seem much more challenging in interactions across species.		# for \(r \to r_1\) from below, cooperation goes extinct in both layers simultaneously following the directed percolation universality class. Surprisingly, the extinction threshold \(r_1\approx0.02283(3)\) is very similar to the threshold in single species interactions \(r\approx0.0261\) (estimated from [[#References\|Szabó & Hauert, 2002]]). Even though conditions for cooperation would seem much more challenging in interactions across species.

	# spontaneous symmetry-breaking arises in the frequency of cooperation on the two layers when \(r \to r_2\). In fact, two equivalent asymmetric phases exist where individuals in one layer prey on those in the other layer or vice versa. This phase transition is similar to the spontaneous symmetry breaking in the two-dimensional Ising model in the absence of an external magnetic field. Although, the additional degrees of freedom (four instead of two microscopic states) result in deviations in the fluctuations of cooperation frequencies.		# spontaneous symmetry-breaking arises in the frequency of cooperation on the two layers when \(r \to r_2\). In fact, two equivalent asymmetric phases exist where individuals in one layer prey on those in the other layer or vice versa. This phase transition is similar to the spontaneous symmetry breaking in the two-dimensional Ising model in the absence of an external magnetic field. Although, the additional degrees of freedom (four instead of two microscopic states) result in deviations in the fluctuations of cooperation frequencies.

	# The third critical transition occurs for \(r \to r_3\) from above and is driven by bursts of \(DD\) domains. For small \(r\) the magnitude of the symmetry breaking decreases with \(r\), while the magnitude of bursts increases and causes fluctuations to diverge. However, the identification of the universal features of bursts and the slow transition towards absorbing states requires further, time-consuming simulations and analytical investigations.		# The third critical transition occurs for \(r \to r_3\) from above and is driven by bursts of \(DD\) domains. For small \(r\) the magnitude of the symmetry breaking decreases with \(r\), while the magnitude of bursts increases and causes fluctuations to diverge. However, the identification of the universal features of bursts and the slow transition towards absorbing states requires further, time-consuming simulations and analytical investigations.

← Older revision		Revision as of 13:54, 10 August 2024
Line 145:		Line 145:
	[[Image:Mutualism - phase transition r1.png\|thumb\|300px\|The scaling of '''a''' \(\bar \rho_a\) and '''b''' \(\chi_a\) versus \((r_1-r)/r_1\) follows the directed percolation universality class (solid lines indicate theoretical expectations).]]		[[Image:Mutualism - phase transition r1.png\|thumb\|300px\|The scaling of '''a''' \(\bar \rho_a\) and '''b''' \(\chi_a\) versus \((r_1-r)/r_1\) follows the directed percolation universality class (solid lines indicate theoretical expectations).]]

	Near the extinction threshold, \(r_1\), isolated clusters of cooperators perform a branching and annihilating random walk with almost perfect correlations between lattices. The characteristic spatial patterns are captured by snapshots of the configuration in each lattice as well as the pair distributions (see [[#Symmetric cooperation: random walk\| Symmetric cooperation: random walk]]).		Near the extinction threshold, \(r_1\), isolated clusters of cooperators perform a branching and annihilating random walk with almost perfect correlations between lattices. The characteristic spatial patterns are captured by snapshots of the configuration in each lattice as well as the pair distributions (see [[#Symmetric cooperation: random walk\|Symmetric cooperation: random walk]]).

	Coordination between lattices is almost perfect with very few \(CD\) and \(DC\) pairs. This is of the essence for the survival of cooperation. \(CC\) interactions are the bulwark against, or at least help to compensate for exploitation by defectors. Interestingly, the emerging spatial configurations are very similar to those in the intra-species donation game on a single lattice.		Coordination between lattices is almost perfect with very few \(CD\) and \(DC\) pairs. This is of the essence for the survival of cooperation. \(CC\) interactions are the bulwark against, or at least help to compensate for exploitation by defectors. Interestingly, the emerging spatial configurations are very similar to those in the intra-species donation game on a single lattice.
Line 151:		Line 151:
	In agreement with theoretical expectations the critical transition at \(r_1\) clearly belongs to the two-dimensional directed percolation universality class: in both layers the frequency of cooperation scales with \(\bar\rho_a \propto (r_{1}-r)^\beta\) for \(\beta=0.580(3)\) and their fluctuations scale with \(\chi_a \propto (r_{1}-r)^{-\gamma}\) for \(\gamma=0.35(1)\) when \(r\to r_1\) from below (as shown on the right)		In agreement with theoretical expectations the critical transition at \(r_1\) clearly belongs to the two-dimensional directed percolation universality class: in both layers the frequency of cooperation scales with \(\bar\rho_a \propto (r_{1}-r)^\beta\) for \(\beta=0.580(3)\) and their fluctuations scale with \(\chi_a \propto (r_{1}-r)^{-\gamma}\) for \(\gamma=0.35(1)\) when \(r\to r_1\) from below (as shown on the right)

	When lowering the cost-to-benefit ratio, \(r_2<r<r_1\), the frequency of cooperation gradually increases with decreasing \(r\), while the coordination between layers decreases. For \(r\) close to \(r_2\) the frequency of all four strategy pairs are essentially the same, which supports that configurations are largely uncorrelated.		When lowering the cost-to-benefit ratio, \(r_2<r<r_1\), the frequency of cooperation gradually increases with decreasing \(r\), while the coordination between layers decreases. For \(r\) close to \(r_2\) the frequency of all four strategy pairs are essentially the same, which supports that configurations are largely uncorrelated (see [[#Symmetric cooperation: uncorrelated\|Symmetric cooperation: uncorrelated]]).

	{{-}}		{{-}}
Line 158:		Line 158:
	[[Image:Mutualism - phase transition r2.png\|thumb\|300px\|The scaling of the order parameter \(\Phi=\vert\bar\rho_{+1}-\bar\rho_{-1}\vert\) as a function of \((r_2-r)/r_2\) is in line with the universality class of spontaneous magnetization in the Ising model. '''b''' The scaling of the fluctuations of \(\Phi\) (\(\bullet\) and \(\circ\)) as well as \(\chi_a\) (\(\Box\) and \(\times\)) above and below the critical threshold \(r_2\). The \(\Phi\) data reproduces the theoretical expectations (solid lines with slopes \(1/8\) in '''a''' and \(-7/4\) in '''b''') within statistical errors (omitted for clarity).]]		[[Image:Mutualism - phase transition r2.png\|thumb\|300px\|The scaling of the order parameter \(\Phi=\vert\bar\rho_{+1}-\bar\rho_{-1}\vert\) as a function of \((r_2-r)/r_2\) is in line with the universality class of spontaneous magnetization in the Ising model. '''b''' The scaling of the fluctuations of \(\Phi\) (\(\bullet\) and \(\circ\)) as well as \(\chi_a\) (\(\Box\) and \(\times\)) above and below the critical threshold \(r_2\). The \(\Phi\) data reproduces the theoretical expectations (solid lines with slopes \(1/8\) in '''a''' and \(-7/4\) in '''b''') within statistical errors (omitted for clarity).]]

	Intriguingly, lowering \(r\) further to \(r_3<r<r_2\) with \(r_3\approx0.00041(3)\) results in spontaneous symmetry breaking with different frequencies in each layer.		Intriguingly, lowering \(r\) further to \(r_3<r<r_2\) with \(r_3\approx0.00041(3)\) results in spontaneous symmetry breaking with different frequencies in each layer (see [[#Asymmetric cooperation\|Asymmetric cooperation]]).

	The distribution of cooperators and defectors in each layer is almost complementary. Clusters, or regions of cooperators, in one layer are matched by defectors in the other. As a consequence, the lattices consist mostly of \(CD\) and \(DC\) pairs. Because of the distinctly different frequencies of cooperation in each lattice, either \(CD\) or \(DC\) pairs dominate. However, note that which pair dominates is of no consequence, because species interactions are symmetric, and hence it is merely a consequence of which species is labelled \(X\) and which \(Y\). Defectors in one species manage to exploit and take advantage of cooperators in the other species. However, exploitation is ephemeral and the scale is limited both spatially as well as temporarily.		The distribution of cooperators and defectors in each layer is almost complementary. Clusters, or regions of cooperators, in one layer are matched by defectors in the other. As a consequence, the lattices consist mostly of \(CD\) and \(DC\) pairs. Because of the distinctly different frequencies of cooperation in each lattice, either \(CD\) or \(DC\) pairs dominate. However, note that which pair dominates is of no consequence, because species interactions are symmetric, and hence it is merely a consequence of which species is labelled \(X\) and which \(Y\). Defectors in one species manage to exploit and take advantage of cooperators in the other species. However, exploitation is ephemeral and the scale is limited both spatially as well as temporarily.
Line 172:		Line 172:
	==== Bursts of defection ====		==== Bursts of defection ====
	[[Image:Mutualism - phase transition r3.png\|thumb\|150px\|<div style="width:296px">Scaling of \(\chi_a\) in each layer (\(\Box\) and \(\times\)) versus \((r-r_3)/r_3\), which is driven by bursts of defection. For reference the solid line shows a power law with exponent \(-1.33\).</div>]]		[[Image:Mutualism - phase transition r3.png\|thumb\|150px\|<div style="width:296px">Scaling of \(\chi_a\) in each layer (\(\Box\) and \(\times\)) versus \((r-r_3)/r_3\), which is driven by bursts of defection. For reference the solid line shows a power law with exponent \(-1.33\).</div>]]

			Finally, for \(0<r<r_3\), the population relaxes into one of its four absorbing states depending on the initial configuration. In fact, the most striking dynamical phenomenon of this model manifests itself in the diverging fluctuations of \(C\) frequencies, \(\chi_a\), for \(r\to r_3\) (see [[#Bursts of defection\|Bursts of defection]]).

			In particular, no theoretical arguments exist to date that explain and justify the observed power law behaviour of \(\chi_a \propto (r-r_3)^{-\gamma}\) for \(r\to r_3\) from above with \(\gamma=1.33(5)\). The divergence of \(\chi_a\) is related to the emergence, growth and elimination of homogeneous \(DD\) domains. Spikes in \(DD\) frequencies drive the dynamics of bursts.

			This type of bursts are driven by sufficiently large spatial domains of an unstable phase. Evidently, their extension and frequency depends on \(r\) and other parameters. In particular, bursts become larger and rarer when decreasing \(r\). Once a burst reaches sizes comparable to the entire lattice, it can drive the evolution towards absorbing states. Interestingly, the absorbing \(DD\) state is rarely observed because it is prone to invasion by any remaining \(CC\) islands, which then pave the way for \(CD\) or \(DC\) phases to grow and take over. This is in stark contrast to intra-species models where cooperation continuously increases for decreasing \(r\), that is as the social dilemma becomes more benign.

			Another peculiarity of the burst dynamics is the feedback between their creation and extinction and the spatial configuration of the asymmetric phase: for \(r\) close to \(r_3\), the asymmetry of cooperation, \(\Phi = \|\bar\rho_{+1} - \bar\rho_{-1}\|\), decreases with \(r\) while the burst sizes increase.

			For \(r<r_3\) the population tends to get absorbed in the fully asymmetric \(CD\) or \(DC\) states. Once one layer is homogeneous the dynamics in the other layer is dominated by neutral drift because the difference in payoffs between cooperators and defectors merely amounts to \(4r\). Hence, for small \(r\), the probability to update the strategy essentially reduces to a coin toss with a slight bias in favour of defectors. The probability to fixate in one or the other absorbing state is thus essentially given by the frequency of strategies in the heterogeneous layer at the time the other layer turned homogeneous. For example, if one layer becomes homogeneous in \(C\), then the frequency of \(C\)’s in the other layer is likely \(<1/2\), due to the asymmetry. Thus, the \(CD\) absorbing state is more likely than \(CC\). Conversely, if one layer becomes homogeneous in $D$, then the frequency of \(C\)’s in the other layer is likely \(>1/2\) and hence the \(DC\) absorbing state is more likely than \(DD\). Finally, it is more likely that one layer becomes homogeneous in \(D\) because isolated \(C\)’s (or small \(C\) clusters) are more easily eliminated than \(D\)’s. Thus, even though all four absorbing states may be reached, in principle, the probabilities to do so are very different and most likely the populations approach heterogeneous absorbing states, whereas homogeneous cooperation across lattices is a highly unlikely outcome.

	{{-}}		{{-}}

← Older revision		Revision as of 09:18, 11 August 2024
Line 162:		Line 162:
	[[Image:Mutualism - phase transition r1.png\|thumb\|300px\|The scaling of '''a''' \(\bar \rho_a\) and '''b''' \(\chi_a\) versus \((r_1-r)/r_1\) follows the directed percolation universality class (solid lines indicate theoretical expectations).]]		[[Image:Mutualism - phase transition r1.png\|thumb\|300px\|The scaling of '''a''' \(\bar \rho_a\) and '''b''' \(\chi_a\) versus \((r_1-r)/r_1\) follows the directed percolation universality class (solid lines indicate theoretical expectations).]]

	Near the extinction threshold, \(r_1\), isolated clusters of cooperators perform a branching and annihilating random walk with almost perfect correlations between lattices. The characteristic spatial patterns are captured by snapshots of the configuration in each lattice as well as the pair distributions (see [[~~#Symmetric cooperation~~: random walk\|Symmetric cooperation: random walk]]).		Near the extinction threshold, \(r_1\), isolated clusters of cooperators perform a branching and annihilating random walk with almost perfect correlations between lattices. The characteristic spatial patterns are captured by snapshots of the configuration in each lattice as well as the pair distributions (see lab on [[EvoLudoLab: Mutualisms - Symmetry, random walk\|Symmetric cooperation: random walk]]).

	Coordination between lattices is almost perfect with very few \(CD\) and \(DC\) pairs. This is of the essence for the survival of cooperation. \(CC\) interactions are the bulwark against, or at least help to compensate for exploitation by defectors. Interestingly, the emerging spatial configurations are very similar to those in the intra-species donation game on a single lattice.		Coordination between lattices is almost perfect with very few \(CD\) and \(DC\) pairs. This is of the essence for the survival of cooperation. \(CC\) interactions are the bulwark against, or at least help to compensate for exploitation by defectors. Interestingly, the emerging spatial configurations are very similar to those in the intra-species donation game on a single lattice.
Line 168:		Line 168:
	In agreement with theoretical expectations the critical transition at \(r_1\) clearly belongs to the two-dimensional directed percolation universality class: in both layers the frequency of cooperation scales with \(\bar\rho_a \propto (r_{1}-r)^\beta\) for \(\beta=0.580(3)\) and their fluctuations scale with \(\chi_a \propto (r_{1}-r)^{-\gamma}\) for \(\gamma=0.35(1)\) when \(r\to r_1\) from below (as shown on the right)		In agreement with theoretical expectations the critical transition at \(r_1\) clearly belongs to the two-dimensional directed percolation universality class: in both layers the frequency of cooperation scales with \(\bar\rho_a \propto (r_{1}-r)^\beta\) for \(\beta=0.580(3)\) and their fluctuations scale with \(\chi_a \propto (r_{1}-r)^{-\gamma}\) for \(\gamma=0.35(1)\) when \(r\to r_1\) from below (as shown on the right)

	When lowering the cost-to-benefit ratio, \(r_2<r<r_1\), the frequency of cooperation gradually increases with decreasing \(r\), while the coordination between layers decreases. For \(r\) close to \(r_2\) the frequency of all four strategy pairs are essentially the same, which supports that configurations are largely uncorrelated (see [[~~#Symmetric cooperation~~: uncorrelated\|Symmetric cooperation: uncorrelated]]).		When lowering the cost-to-benefit ratio, \(r_2<r<r_1\), the frequency of cooperation gradually increases with decreasing \(r\), while the coordination between layers decreases. For \(r\) close to \(r_2\) the frequency of all four strategy pairs are essentially the same, which supports that configurations are largely uncorrelated (see lab on [[EvoLudoLab: Mutualisms - Symmetry, uncorrelated\|Symmetric cooperation: uncorrelated]]).

	{{-}}		{{-}}

Line 175:		Line 174:
	[[Image:Mutualism - phase transition r2.png\|thumb\|300px\|The scaling of the order parameter \(\Phi=\vert\bar\rho_{+1}-\bar\rho_{-1}\vert\) as a function of \((r_2-r)/r_2\) is in line with the universality class of spontaneous magnetization in the Ising model. '''b''' The scaling of the fluctuations of \(\Phi\) (\(\bullet\) and \(\circ\)) as well as \(\chi_a\) (\(\Box\) and \(\times\)) above and below the critical threshold \(r_2\). The \(\Phi\) data reproduces the theoretical expectations (solid lines with slopes \(1/8\) in '''a''' and \(-7/4\) in '''b''') within statistical errors (omitted for clarity).]]		[[Image:Mutualism - phase transition r2.png\|thumb\|300px\|The scaling of the order parameter \(\Phi=\vert\bar\rho_{+1}-\bar\rho_{-1}\vert\) as a function of \((r_2-r)/r_2\) is in line with the universality class of spontaneous magnetization in the Ising model. '''b''' The scaling of the fluctuations of \(\Phi\) (\(\bullet\) and \(\circ\)) as well as \(\chi_a\) (\(\Box\) and \(\times\)) above and below the critical threshold \(r_2\). The \(\Phi\) data reproduces the theoretical expectations (solid lines with slopes \(1/8\) in '''a''' and \(-7/4\) in '''b''') within statistical errors (omitted for clarity).]]

	Intriguingly, lowering \(r\) further to \(r_3<r<r_2\) with \(r_3\approx0.00041(3)\) results in spontaneous symmetry breaking with different frequencies in each layer (see [[#Asymmetric cooperation\|Asymmetric cooperation]]).		Intriguingly, lowering \(r\) further to \(r_3<r<r_2\) with \(r_3\approx0.00041(3)\) results in spontaneous symmetry breaking with different frequencies in each layer (see lab on [[EvoLudoLab: Mutualisms - Asymmetric cooperation\|Asymmetric cooperation]]).

	The distribution of cooperators and defectors in each layer is almost complementary. Clusters, or regions of cooperators, in one layer are matched by defectors in the other. As a consequence, the lattices consist mostly of \(CD\) and \(DC\) pairs. Because of the distinctly different frequencies of cooperation in each lattice, either \(CD\) or \(DC\) pairs dominate. However, note that which pair dominates is of no consequence, because species interactions are symmetric, and hence it is merely a consequence of which species is labelled \(X\) and which \(Y\). Defectors in one species manage to exploit and take advantage of cooperators in the other species. However, exploitation is ephemeral and the scale is limited both spatially as well as temporarily.		The distribution of cooperators and defectors in each layer is almost complementary. Clusters, or regions of cooperators, in one layer are matched by defectors in the other. As a consequence, the lattices consist mostly of \(CD\) and \(DC\) pairs. Because of the distinctly different frequencies of cooperation in each lattice, either \(CD\) or \(DC\) pairs dominate. However, note that which pair dominates is of no consequence, because species interactions are symmetric, and hence it is merely a consequence of which species is labelled \(X\) and which \(Y\). Defectors in one species manage to exploit and take advantage of cooperators in the other species. However, exploitation is ephemeral and the scale is limited both spatially as well as temporarily.
Line 190:		Line 189:
	[[Image:Mutualism - phase transition r3.png\|thumb\|150px\|<div style="width:296px">Scaling of \(\chi_a\) in each layer (\(\Box\) and \(\times\)) versus \((r-r_3)/r_3\), which is driven by bursts of defection. For reference the solid line shows a power law with exponent \(-1.33\).</div>]]		[[Image:Mutualism - phase transition r3.png\|thumb\|150px\|<div style="width:296px">Scaling of \(\chi_a\) in each layer (\(\Box\) and \(\times\)) versus \((r-r_3)/r_3\), which is driven by bursts of defection. For reference the solid line shows a power law with exponent \(-1.33\).</div>]]

	Finally, for \(0<r<r_3\), the population relaxes into one of its four absorbing states depending on the initial configuration. In fact, the most striking dynamical phenomenon of this model manifests itself in the diverging fluctuations of \(C\) frequencies, \(\chi_a\), for \(r\to r_3\) (see [[#Bursts of defection\|Bursts of defection]]).		Finally, for \(0<r<r_3\), the population relaxes into one of its four absorbing states depending on the initial configuration. In fact, the most striking dynamical phenomenon of this model manifests itself in the diverging fluctuations of \(C\) frequencies, \(\chi_a\), for \(r\to r_3\) (see lab on [[EvoLudoLab: Mutualisms - Bursts of defection\| Bursts of defection]]).

	In particular, no theoretical arguments exist to date that explain and justify the observed power law behaviour of \(\chi_a \propto (r-r_3)^{-\gamma}\) for \(r\to r_3\) from above with \(\gamma=1.33(5)\). The divergence of \(\chi_a\) is related to the emergence, growth and elimination of homogeneous \(DD\) domains. Spikes in \(DD\) frequencies drive the dynamics of bursts.		In particular, no theoretical arguments exist to date that explain and justify the observed power law behaviour of \(\chi_a \propto (r-r_3)^{-\gamma}\) for \(r\to r_3\) from above with \(\gamma=1.33(5)\). The divergence of \(\chi_a\) is related to the emergence, growth and elimination of homogeneous \(DD\) domains. Spikes in \(DD\) frequencies drive the dynamics of bursts.