Introspection Dynamics with Mutation in Additive Games
Pith reviewed 2026-05-25 02:48 UTC · model grok-4.3
The pith
Introspection dynamics yield exact closed-form long-run cooperation probability in heterogeneous public goods games.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
Under the state-independence condition the introspection Markov chain decomposes as a random-scan product of N independent two-state chains and the stationary distribution is a product measure; for the heterogeneous public goods game the long-run cooperation probability admits the exact closed form p_C = (1/N) sum_i [(1-μ_i0-μ_i1)/(1+exp(β_i α_i (1-r_i/N))) + μ_i0].
What carries the argument
State-independence of the payoff difference Δf_i, which enables the Markov chain to factor into N independent two-state chains whose product measure gives the stationary distribution.
Load-bearing premise
The payoff difference evaluated by each player when considering a switch depends only on their own action and not on the current actions of the others.
What would settle it
For small N and chosen parameters, run long simulations of the full Markov chain and verify whether the time-averaged fraction of cooperators converges to the value given by the closed-form sum.
Figures
read the original abstract
Cooperation in heterogeneous groups, where individuals differ in resources, productivity, and behavioural responsiveness, underpins collective action across social and biological systems. Introspection dynamics, in which each player compares their payoff to their payoff under the alternative action, provides a natural learning rule for such asymmetric settings. Couto and Pal showed that for additive games, those in which the payoff difference a player evaluates when considering a switch is independent of the other players' actions, the stationary distribution of introspection dynamics is a product measure. We extend this result to introspection dynamics with mutation, where a selected player switches to a random action with some probability independent of payoffs, and with player-specific selection intensities. We show that the product structure is preserved, and we obtain the explicit per-player cooperation probability $p_i=\phi_i(\delta_i)(1-\mu_{i0}-\mu_{i1})+\mu_{i0}$. We consider the heterogeneous public goods game, where $N$ players may differ in their contributions $\alpha_i$, public goods multipliers $r_i$, and selection intensities $\beta_i$; the long-run cooperation probability admits the closed form $$ p_C = \frac{1}{N}\sum_{i=1}^{N} \left[\frac{1-\mu_{i0}-\mu_{i1}}{1+e^{\,\beta_i\alpha_i(1-r_i/N)}}+\mu_{i0}\right]. $$ Several structural consequences follow: a player-specific cooperation threshold at $r_i = N$ under symmetric mutation, a neutral-drift regime in which cooperation is governed entirely by mutation bias, and a mutation-selection balance in which aggregate cooperation is affine in the mutation rate, interpolating between the selection-driven level and neutrality. Mutation also regularises the strong-selection limit, so the closed form holds as $\beta_i\to\infty$, where the mutation-free dynamics degenerate.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript claims that introspection dynamics on multiplayer games with state-independent payoff differences Δf_i admit an exact Markov-chain decomposition: the random-scan chain factors as a product of N independent two-state chains whose stationary distribution is a product measure. For the heterogeneous public goods game the linear payoff structure f_i = −α_i s_i + (r_i/N) ∑_j α_j s_j implies this state-independence, yielding the closed-form stationary cooperation probability p_C = (1/N) ∑_i [(1−μ_i0−μ_i1)/(1+exp(β_i α_i (1−r_i/N))) + μ_i0] with no approximation.
Significance. If the result holds, the exact closed-form expression is a notable contribution to evolutionary game theory, where stationary distributions for heterogeneous populations are typically obtained only via simulation or mean-field approximations. The decomposition into independent marginals and the immediate structural corollaries (player-specific threshold at r_i = N, payoff neutrality at β_i = 0) follow directly from the linear payoff assumption and constitute a clean, falsifiable prediction.
minor comments (2)
- [Markov-chain decomposition paragraph] The random-scan update schedule is stated but not illustrated; a one-paragraph example with N=2 would clarify the product structure for readers.
- [Equation for p_C] The Fermi factor is written without an explicit reference to the standard form used in the literature; adding a short parenthetical citation would aid accessibility.
Simulated Author's Rebuttal
We thank the referee for their positive evaluation of the manuscript and their recommendation to accept. The report accurately captures the main technical contribution: the state-independence property of introspection dynamics under linear payoffs, the resulting Markov-chain decomposition, and the exact closed-form expression for the stationary cooperation probability in the heterogeneous public goods game.
Circularity Check
No significant circularity identified
full rationale
The paper derives the closed-form p_C directly from the linear payoff structure f_i = −α_i s_i + (r_i/N) ∑_j α_j s_j, which makes Δf_i independent of other players' states by algebraic cancellation of all cross terms. This state-independence allows the Markov chain to factor exactly into N independent two-state chains whose stationary marginals are written down explicitly; the given product-measure average is the immediate consequence of that factorization plus the standard Fermi update and mutation terms. No parameter is fitted to data and then relabeled as a prediction, no ansatz is smuggled via self-citation, and the uniqueness of the decomposition follows from the random-scan construction rather than from any prior result by the same authors. The derivation is therefore self-contained against the stated assumptions.
Axiom & Free-Parameter Ledger
free parameters (1)
- mutation probabilities μ_i0 and μ_i1
axioms (2)
- standard math Finite irreducible Markov chain possesses a unique stationary distribution
- domain assumption Payoff difference Δf_i is independent of other players' actions (state-independence)
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.