Independent scholarship · Open accessThe monthly digest

Indexed text attachment

evolutionary-game-theory-in-higher-categories.tex

Source for Evolutionary Game Theory in Higher Categories

View paper · Download original file

Lines 369–468 of 3949. Source text is displayed literally.

369\[370 x_{\varepsilon}=(1-\varepsilon)p+\varepsilon q.371\]372The pair \(p,q\) may consist of mixed strategies. The373parameter \(\varepsilon\) specifies the share of the374competitor in the population. It does not describe the375probabilities within either of those two strategies.376Define the payoff advantage in the resulting population by377\[378 D_{p,q}(\varepsilon)379   =u_A(p,x_{\varepsilon})-u_A(q,x_{\varepsilon}).380\]381The resident \(p\) is evolutionarily stable if for each382\(q\ne p\) there exists a threshold \(b(q)>0\) such that383\(D_{p,q}(\varepsilon)>0\) whenever384\(0<\varepsilon<\min\{1,b(q)\}\).385The threshold in this definition can depend on \(q\).386A threshold valid for all competitors is a different387quantifier placement, addressed in388Chapter~\ref{ch:superiority}.389 390\begin{theorem}[Invasion criterion for pairwise games]391\label{thm:invasion-criterion}392A resident \(p\) is evolutionarily stable if and only393if it is symmetric Nash and, for each \(q\ne p\) with394\(u_A(p,p)=u_A(q,p)\), one has395\[396 u_A(p,q)>u_A(q,q).397\]398\end{theorem}399 400\begin{proof}401For each fixed competitor write402\(d_0=u_A(p,p)-u_A(q,p)\) and403\(d_1=u_A(p,q)-u_A(q,q)\). The affine law in the opponent404argument gives405\[406 D_{p,q}(\varepsilon)=(1-\varepsilon)d_0+\varepsilon d_1 .407\]408If this is positive on a sufficiently small interval of409positive \(\varepsilon\), continuity gives \(d_0\geq0\).410When \(d_0=0\), the same interval condition requires411\(d_1>0\) by the displayed identity. Conversely, if412\(d_0>0\), the advantage remains positive for all413sufficiently small \(\varepsilon\) by continuity. If414\(d_0=0\) and \(d_1>0\), the advantage is415\(\varepsilon d_1>0\) throughout the range of positive416mutant shares. Applying the argument to every \(q\ne p\)417gives the equivalence.418\end{proof}419 420The two terms are evaluated in different environments.421The first evaluates both strategies against a purely422resident population. The second evaluates both against423the competing strategy. In the event of a first-order424tie, the strictly positive second comparison explains425why the resident gains a strict advantage when a426positive mutant share is present. Every strict symmetric427Nash equilibrium is evolutionarily stable, since none428of its competitors ties in the first comparison.429 430For a general continuous fitness map the first comparison431still gives a necessary condition for positive advantage432near zero mutant share. The exact tie test in433Theorem~\ref{thm:invasion-criterion} uses affine dependence434on the population. With nonlinear dependence, the advantage435along the segment can have a nontrivial shape. An endpoint436comparison then requires additional hypotheses before it437determines the sign near the resident.438 439\section{Neutral stability and sets of solutions}440 441Neutral stability uses the weak inequality442\(D_{p,q}(\varepsilon)\geq0\) in the invasion test. This443standard variant allows a competing strategy to match the444resident's score when its share is small445\cite[Ch.~2]{weibull}. The same proof, with a weak inequality446on the tie condition, shows that \(p\) is neutrally stable447exactly when it is symmetric Nash and \(d_1\geq0\) for each448distinct \(q\) with \(d_0=0\). Thus the solution predicates449satisfy450\[451 \operatorname{ESS}(A)\subseteq452 \operatorname{NS}(A)\subseteq453 \operatorname{Nash}(A).454\]455These are subsets of the state space. An equality between456local payoffs in the invasion test need not select one457of the tied states. Persistence or attraction of the458resident depends on the dynamics chosen to resolve or459retain such ties.460 461The symmetric Nash condition defines a closed relation462between payoff matrices and population states. Indeed,463for each fixed pure type \(i\) it is the nonnegativity464of the continuous difference \(u_A(p,p)-(Ap)_i\), and465the pure inequalities imply the conditions for all466mixtures by averaging. An evolutionary stability467relation adds a conditional strict inequality. For468instance, along the parameter ray \(A(t)=tA\) for