Independent scholarship · Open accessThe monthly digest

Indexed text attachment

games-and-information-in-higher-categories.tex

Source for Games and Information in Higher Categories

View paper · Download original file

Lines 3041–3140 of 6559. Source text is displayed literally.

3041  winning strategy.3042\item In the discounted repeated Prisoner's Dilemma there are computable strategies to which a best response3043  exists but no best response is computable.3044\end{enumerate}3045\end{theorem}3046 3047The proofs are diagonal: a strategy can be built to consult what a given machine would do and to punish it,3048so that the task of responding optimally embeds an undecidable problem. A player modelled as a Turing3049machine cannot in general predict a player like herself \cite{Bin87}. This is the third appearance of the3050diagonal in this book---after the nonexistence of a universal possibility structure3051(\cref{prop:no-universal-possibility}) and alongside the non-constructive character of Nash's theorem---and3052it marks a real boundary: the equilibrium concept quantifies over the whole terminal coalgebra, while3053agents inhabit its computable part.3054 3055\section*{Notes}3056 3057Standard references for repeated games are \cite{MS06,FT91}. The chain-store paradox is \cite{Sel78}. The3058coalgebraic view of automata is expounded in \cite{Rut00}; automata as strategies were introduced by3059Rubinstein \cite{Rub86} and their complexity analysed in \cite{KS88,Ney85}. The one-shot deviation3060principle descends from Blackwell's dynamic programming \cite{Bla65}. Coinductive treatments of infinite3061games and their equilibria, including cases without discounting, are \cite{LP12,AW17}; the infinite product3062of selection functions of \cref{sec:selection} is another. The topos of trees and its guarded fixed points3063are from \cite{BMSS12}. Self-generation is due to Abreu, Pearce and Stacchetti \cite{APS90}, building on3064\cite{Abr88}. The folk theorem with Nash reversion is Friedman's \cite{Fri71}; the general3065perfect-monitoring result is Fudenberg and Maskin's \cite{FM86}. Markov perfect equilibrium is defined in3066general in \cite{MT01}. Evolutionary stability is from \cite{MSP73}. Computability limits on strategic3067reasoning are discussed in \cite{Rab57,Bin87,Kno94,NZ96}.3068 3069\section*{Exercises}3070 3071\begin{exercise}3072Write tit-for-tat and ``win-stay, lose-shift'' as coalgebras for $F(X)=A_i\times X^A$ and compute the3073unique maps to the terminal coalgebra on the histories of length at most two.3074\end{exercise}3075\begin{exercise}3076In the Prisoner's Dilemma of \cref{ex:grim}, for which $\delta$ is the profile ``both play tit-for-tat'' a3077Nash equilibrium? Show that it is not subgame perfect for generic $\delta$, by exhibiting a history at which a one-shot3078deviation is profitable.3079\end{exercise}3080\begin{exercise}3081Compute $B(W)$ for $W=\{(1,1)\}$ and for $W=\{(2,2)\}$ in the Prisoner's Dilemma, and conclude that3082$\{(2,2)\}$ is not self-generating for any $\delta$.3083\end{exercise}3084\begin{exercise}3085Show that the operator $B$ is monotone, and give an example of a stage game and a $\delta$ for which3086$\bigcup_nB^n(\emptyset)\neq\bigcap_nB^n(W_0)$.3087\end{exercise}3088\begin{exercise}3089Consider the entry game of \cref{ex:entry} repeated infinitely against a sequence of short-lived entrants,3090with the incumbent discounting at $\delta$. Find the least $\delta$ for which there is a subgame perfect3091equilibrium with no entry on the path.3092\end{exercise}3093\begin{exercise}3094Verify that in Hawk--Dove the strategy ``hawk with probability $V/C$'' satisfies the definition of3095evolutionary stability, and compute the mean fitness of the population at that point.3096\end{exercise}3097%% ---- end ch/09-repetition.tex ----3098%% ---- begin ch/10-dynamic.tex ----3099\chapter{Dynamic Games of Incomplete Information}\label{ch:dynamic}3100 3101When a player moves at an information set that is not the root of a subgame, subgame perfection has3102nothing to say about her, and yet she may be making an incredible threat. To test her choice one must say3103what she believes about where she is. This chapter combines the temporal modality of \cref{ch:perfection}3104with the beliefs of \cref{ch:belief}; studies the geometry of beliefs at events of probability zero; and3105then shows, through reputation and the electronic mail game, how violently equilibrium can respond to3106small changes in what is believed and known.3107 3108\section{Sequential rationality}3109 3110\begin{definition}\Scl[\cite{KW82a}]3111An \term{assessment} in an extensive game is a pair $(\sigma,\mu)$ of a behavioural strategy profile and a3112\emph{system of beliefs}: for each information set $I$ a distribution $\mu(I)$ over the nodes of $I$. It is3113\term{sequentially rational} if at every information set $I$ of every player $i$, the continuation of $\sigma_i$3114maximizes $i$'s expected payoff given $\mu(I)$ and $\sigma_{-i}$.3115\end{definition}3116 3117A system of beliefs is a section of the bundle $\coprod_I\Dist(I)\to\mathcal{I}$, just as a strategy is a3118section of the bundle of actions. Sequential rationality is the necessity modality of \cref{prop:box} with3119subgames replaced by information sets, each equipped with the belief that turns it into a decision3120problem: perfection and knowledge, composed.3121 3122\begin{example}[Where subgame perfection is silent]\label{ex:seq-entry}3123An entrant chooses \emph{Out}, with payoffs $(0,4)$, or enters at a \emph{small} or a \emph{large} scale. The3124incumbent observes entry but not its scale, and fights or accommodates. Accommodating small entry gives3125$(1,2)$, fighting it $(-1,3)$; accommodating large entry gives $(2,1)$, fighting it $(-2,-1)$. There is no3126proper subgame. Let $m$ be the incumbent's belief that entry is small. Fighting is optimal iff3127$3m-(1-m)\ge2m+(1-m)$, that is $m\ge\tfrac23$. There are two kinds of sequentially rational assessment: the3128entrant enters large, the incumbent accommodates, $m=0$; and the entrant stays out, the incumbent would3129fight, sustained by any belief $m\ge\tfrac23$ at an information set that is never reached. Had fighting3130been worse than accommodating against both scales, no belief would support it, and the second equilibrium3131would disappear although it remains subgame perfect.3132\end{example}3133 3134On the equilibrium path beliefs are determined by Bayes' rule. Off the path, the conditioning event has3135probability zero and Bayes' rule is silent; in the example any $m$ is allowed. The various definitions of3136\term{perfect Bayesian equilibrium} \cite{FT91b} add requirements of the form ``update by Bayes' rule wherever3137possible''. Kreps and Wilson's definition is the cleanest.3138 3139\begin{definition}\Scl[\cite{KW82a}]3140An assessment $(\sigma,\mu)$ is \term{consistent} if it is the limit of assessments $(\sigma^k,\mu^k)$ in