content/applied-modal-logic/temporal-logic/temporal-logic.tex
1% Part: applied-modal-logics2% Chapter: temporal-logic34\documentclass[../../../include/open-logic-chapter]{subfiles}56\begin{document}78\olchapter{aml}{tl}{Temporal Logics}910\begin{editorial}11 This chapter covers temporal logics.12\end{editorial}1314\olimport{introduction}15\olimport{temporal-logic-semantics}16\olimport{properties-accessibility}17\olimport{extra-temporal-operators}18\olimport{possible-histories}1920\OLEndPartHook2122\end{document}
content/applied-modal-logic/temporal-logic/introduction.tex
1% Part: applied-modal-logics2% Chapter: temporal-logic3% Section: introduction45\documentclass[../../../include/open-logic-section]{subfiles}67\begin{document}89\olfileid{aml}{tl}{int}1011\olsection{Introduction}1213Temporal logics deal with claims about things that will or have been14the case. Arthur Prior is credited as the originator of temporal15logic, which he called \emph{tense logic}. Our treatment of temporal16logic here will largely follow Prior's original modal treatment of17introducing temporal operators into the basic framework of18propositional logic, which treats claims as generally lacking in19tense.2021For example, in propositional logic, I might talk about a dog, Beezie,22who sometimes sits and sometimes doesn't sit, as dogs are wont to do.23It would be contradictory in classical logic to claim that Beezie is24sitting and also that Beezie is not sitting. But obviously both can be25true, just not at the same time; adding temporal operators to the26language can allow us to express that claim relatively easily. The27addition of temporal operators also allows us to account for the28validity of inferences like the one from ``Beezie will get a treat or29a ball" to ``Beezie will get a treat or Beezie will get a ball." 3031However, a lot of philosophical issues arise with temporal logic that32might lead us to adopt one framework of temporal logic over another.33For example, a future contingent is a statement about the future that34is neither necessary nor impossible. If we say ``Richard will go to the35grocery store tomorrow," we are expressing a claim about something that36has not yet happened, and whose truth value is contestable. In fact,37it is contestable whether that claim can even be \emph{assigned} a38truth value in the first place. If we are strict determinists, then39perhaps we can be comfortable with the idea that this sentence is in40fact true or false, even before the event in question is supposed to41take place---it just may be that we do not know its truth value yet.42In contrast, we might believe in a genuinely open future, in which the43truth values of future contingents are undetermined. 4445As it turns out, a lot of these commitments about the structure and46nature of time are built in to our choices of models and frameworks of47temporal logics. For example, we might ask ourselves whether we should48construct models in which time is linear, branching or even circular.49We might have to make decisions about whether our temporal models will50have beginning and end points, and whether time is to be represented51using discrete instants or as a continuum. 5253\end{document}
content/applied-modal-logic/temporal-logic/temporal-logic-semantics.tex
1% Part: applied-modal-logics2% Chapter: temporal-logic3% Section: language-epistemic-logic45\documentclass[../../../include/open-logic-section]{subfiles}67\begin{document}89\olfileid{aml}{tl}{sem}1011\olsection{Semantics for Temporal Logic}1213\begin{defn}14The basic language of temporal logic contains15\begin{enumerate}16 \tagitem{prvFalse}{The propositional constant for !!{falsity}~$\lfalse$.}{}17 \tagitem{prvTrue}{The propositional constant for !!{truth}~$\ltrue$.}{}18 \item A !!{denumerable}s set of !!{propositional variable}s: $\Obj19 p_0$, $\Obj p_1$, $\Obj p_2$, \dots20 \item The propositional connectives: \startycommalist21 \iftag{prvNot}{\ycomma $\lnot$ (negation)}{}%22 \iftag{prvAnd}{\ycomma $\land$ (conjunction)}{}%23 \iftag{prvOr}{\ycomma $\lor$ (disjunction)}{}%24 \iftag{prvIf}{\ycomma $\lif$ (!!{conditional})}{}%25 \iftag{prvIff}{\ycomma $\liff$ (!!{biconditional})}{}.26 \item Past operators $\Ptemp$ and $\Htemp$.27 \item Future operators $\Ftemp$ and $\Gtemp$.28\end{enumerate}29\end{defn}3031Later on, we will discuss the potential addition of other kinds of modal operators.3233\begin{defn}34\emph{!!^{formula}s} of the temporal language are inductively35 defined as follows:36\begin{enumerate}37\tagitem{prvFalse}{$\lfalse$ is an atomic !!{formula}.}{}3839\tagitem{prvTrue}{$\ltrue$ is an atomic !!{formula}.}{}4041\item Every propositional variable $\Obj p_i$ is an (atomic) !!{formula}.4243\tagitem{prvNot}{If $!A$ is !!a{formula}, then $\lnot !A$ is44 !!a{formula}.}{}4546\tagitem{prvAnd}{If $!A$ and $!B$ are !!{formula}s, then $(!A \land47 !B)$ is !!a{formula}.}{}4849\tagitem{prvOr}{If $!A$ and $!B$ are !!{formula}s, then $(!A \lor !B)$50 is !!a{formula}.}{}5152\tagitem{prvIf}{If $!A$ and $!B$ are !!{formula}s, then $(!A \lif !B)$53 is !!a{formula}.}{}5455\tagitem{prvIff}{If $!A$ and $!B$ are !!{formula}s, then $(!A \liff !B)$56 is !!a{formula}.}{}5758\item If $!A$ is !!a{formula}, then $\Ptemp !A$, $\Htemp !A$, $F !A$, $\Gtemp !A$ are all59 !!{formula}s.6061\tagitem{limitClause}{Nothing else is !!a{formula}.}{}62\end{enumerate}63\end{defn}6465The semantics of temporal logics are given in terms of relational models, as with other kinds of intensional logics. 6667\begin{defn}68 A \emph{model} for temporal language is a triple69 $\mModel{M} = \tuple{T, \prec, V}$, where70 \begin{enumerate}71 \item $T$ is a nonempty set, interpreted as points in time.72 \item $\prec$ is a binary relation on $T$.73 \item $V$ is a function assigning to each !!{propositional74 variable}~$p$ a set $V(p)$ of points in time.75 \end{enumerate}76 When $t \prec t'$ holds, we say that $t$ \emph{precedes}~$t'$. 77 When $t \in V(p)$ we say $p$ is \emph{true at}~$t$.78\end{defn}7980For now, you will notice that we do not impose any conditions on our precedence relation $\prec$. This means that at present, there are no restrictions on the structure of our temporal models, so we could have models in which time is linear, branching, circular, or has any structure whatsoever. 8182Just as with normal modal logic, every temporal model determines which83!!{formula}s count as true at which points in it. We use the same84notation ``model $\mModel{M}$ makes !!{formula}~$!A$ true at85point~$t$'' for the basic notion of relational semantics. The relation86is defined inductively and is identical to the normal modal case for87all non-modal operators.8889\begin{defn}\ollabel{defn:tmodels}90 \emph{Truth of !!a{formula}~$!A$ at~$t$} in a~$\mModel M$, in symbols:91 $\mSat{M}{!A}[t]$, is defined inductively as follows:92 \begin{enumerate}93 \tagitem{prvFalse}{\indcase{!A}{\lfalse}{Never $\mSat{M}{\lfalse}[t]$}.}{}94 \tagitem{prvTrue}{\indcase{!A}{\ltrue}{Always $\mSat{M}{\ltrue}[t]$}.}{}95 \item $\mSat{M}{p}[t]$ iff $t \in V(p)$96 \tagitem{prvNot}{\indcase{!A}{\lnot !B}{$\mSat{M}{\indfrm}[t]$ iff97 $\mSat/{M}{!B}[t]$}.}{}98 \tagitem{prvAnd}{\indcase{!A}{(!B \land !C)}{$\mSat{M}{\indfrm}[t]$ iff99 $\mSat{M}{!B}[t]$ and $\mSat{M}{!C}[t]$}.}{}100 \tagitem{prvOr}{\indcase{!A}{(!B \lor !C)}{$\mSat{M}{\indfrm}[t]$ iff101 $\mSat{M}{!B}[t]$ or $\mSat{M}{!C}[t]$} (or both).}{}102 \tagitem{prvIf}{\indcase{!A}{(!B \lif !C)}{$\mSat{M}{\indfrm}[t]$ iff103 $\mSat/{M}{!B}[t]$ or $\mSat{M}{!C}[t]$}.}{}104 \tagitem{prvIff}{\indcase{!A}{(!B \liff !C)}{$\mSat{M}{\indfrm}[t]$ iff105 either both $\mSat{M}{!B}[t]$ and $\mSat{M}{!C}[t]$ or106 neither $\mSat{M}{!B}[t]$ nor $\mSat{M}{!C}[t]$}.}{}107 \item\ollabel{defn:sub:mmodels-p}108 \indcase{!A}{\Ptemp !B}{$\mSat{M}{\indfrm}[t]$ iff109 $\mSat{M}{!B}[t']$ for some $t' \in T$ with $t' \prec t$}110\item\ollabel{defn:sub:mmodels-h}111 \indcase{!A}{\Htemp !B}{$\mSat{M}{\indfrm}[t]$ iff112 $\mSat{M}{!B}[t']$ for every $t' \in T$ with $t' \prec t$}113 \item\ollabel{defn:sub:mmodels-f}114 \indcase{!A}{\Ftemp !B}{$\mSat{M}{\indfrm}[t]$ iff115 $\mSat{M}{!B}[t']$ for some $t' \in T$ with $t \prec t'$}116\item\ollabel{defn:sub:mmodels-g}117 \indcase{!A}{\Gtemp !B}{$\mSat{M}{\indfrm}[t]$ iff118 $\mSat{M}{!B}[t']$ for every $t' \in T$ with $t \prec t'$}119 \end{enumerate} 120\end{defn}121122Based on the semantics, you might be able to see that the operators $\Ptemp$ and~$\Htemp$ are duals, as well as the123operators $\Ftemp$ and $\Gtemp$, such that we could define $\Htemp !A$ as $\lnot \Ptemp \lnot !A$, and the same with $\Gtemp$ and~$\Ftemp$.124125\end{document}
content/applied-modal-logic/temporal-logic/properties-accessibility.tex
1% Part: applied-modal-logics2% Chapter: temporal-logic3% Section: properties-accessibility45\documentclass[../../../include/open-logic-section]{subfiles}67\begin{document}89\olfileid{aml}{tl}{acc}1011\olsection{Properties of Temporal Frames}1213Given that our temporal models do not impose any conditions on the14relation~$\prec$, the only one of our familiar axioms that holds in15all models is~$K$, or its analogues $K_{\Gtemp}$ and~$K_{\Htemp}$:16\begin{align*}17\tag{$K_{\Gtemp}$} \Gtemp (p \to q) & \to (\Gtemp p \to \Gtemp q)\\18\tag{$K_{\Htemp}$} \Htemp (p \to q) & \to (\Htemp p \to \Htemp q)19\end{align*}2021However, if we want our models to impose stricter conditions on how22time is represented, for instance by ensuring that $\prec$ is a linear23order, then we will end up with other validities in our models.2425\begin{table}[t]26 \begin{tabular}{| p{.48\textwidth} || p{.48\textwidth} |}27 \hline28 {\emph{If $\prec$ is \dots}} & {\emph{then \dots is true in~$\mModel{M}$:}} \\29 \hline \hline30 \emph{transitive}: \newline31 $\forall u \forall v \forall w ((u \prec v \land v \prec w) \lif u \prec w)$ & 32 $\Ftemp \Ftemp p \lif \Ftemp p$ \\33 \hline 34 \emph{linear}: \newline35 $\forall w \forall v (w \prec v \lor w = v \lor v \prec w)$ & 36 $(\Ftemp \Ptemp p \lor \Ptemp \Ftemp p) \lif (\Ptemp p \lor p \lor \Ftemp p)$\\37 \hline38 \emph{dense}: \newline39 $\forall w \forall v (w \prec v \to \exists u(w \prec u \land u \prec v))$ & 40 $\Ftemp p \lif \Ftemp \Ftemp p$ \\41 \hline42 \emph{unbounded (past)}: \newline43 $\forall w \exists v( v \prec w)$ & 44 $\Htemp p \to \Ptemp p$ \\45 \hline46 \emph{unbounded (future)}: \newline47 $\forall w \exists v( w \prec v)$ & 48 $\Gtemp p \to \Ftemp p$ \\49 \hline50 \end{tabular}51 \caption{Some temporal frame correspondence properties.}52 \ollabel{tab:correspondence}53\end{table} 5455Several of the properties from \olref{tab:correspondence} might seem56like desirable features for a model that is intended to represent57time. However, it is worth noting that, even though we can impose58whichever conditions we like on the $\prec$ relation, not all59conditions correspond to !!{formula}s that can be expressed in the60language of temporal logic. For example, irreflexivity, or the idea61that $\forall w \lnot (w \prec w)$, does not have a corresponding62formula in temporal logic. 6364\end{document}
content/applied-modal-logic/temporal-logic/extra-temporal-operators.tex
1% Part: applied-modal-logics2% Chapter: temporal-logic3% Section: extra-operators45\documentclass[../../../include/open-logic-section]{subfiles}67\begin{document}89\olfileid{aml}{tl}{ext}1011\olsection{Additional Operators for Temporal Logic}1213In addition to the unary operators for past and future, temporal14logics also sometimes include binary operators $\Since $ and~$\Until$,15intended to symbolize ``since'' and ``until''. This means adding16$\Since $ and~$\Until$ into the language of temporal logic and adding17the following clause into the definition of a temporal !!{formula}:1819\begin{itemize}20\item[] If $!A$ and $!B$ are !!{formula}s, then $(\Since !A !B)$ and21 $(\Until !A !B)$ are both !!{formula}s.22\end{itemize}2324The semantics for these operators are then given as follows:2526\begin{defn}\ollabel{defn:since-until}27 \emph{Truth of !!a{formula}~$!A$ at~$t$} in a~$\mModel M$:28 \begin{enumerate}29 \item\ollabel{defn:sub:mmodels-since} \indcase{!A}{\Since !B30 !C}{$\mSat{M}{\indfrm}[t]$ iff $\mSat{M}{!B}[t']$ for some $t' \in31 T$ with $t' \prec t$, and for all $s$ with $t' \prec s \prec t$,32 $\mSat{M}{!C}[s]$}33 \item\ollabel{defn:sub:mmodels-until} \indcase{!A}{\Until !B34 !C}{$\mSat{M}{\indfrm}[t]$ iff $\mSat{M}{!B}[t']$ for some $t' \in35 T$ with $t \prec t'$, and for all $s$ with $t \prec s \prec t'$,36 $\mSat{M}{!C}[s]$}37 \end{enumerate} 38\end{defn}3940The intuitive reading of $\Since !B !C$ is ``Since $!B$ was the case,41$!C$~has been the case.'' And the intuitive reading of $\Until !B !C$42is ``Until $!B$ will be the case, $!C$~will be the case.''4344\end{document}
content/applied-modal-logic/temporal-logic/possible-histories.tex
1% Part: applied-modal-logics2% Chapter: temporal-logic3% Section: possible-histories45\documentclass[../../../include/open-logic-section]{subfiles}67\begin{document}89\olfileid{aml}{tl}{poss}1011\olsection{Possible Histories}1213The relational models of temporal logic that we have been using are14extremely flexible, since we do not have to place any restrictions on15the accessibility relation. This means that temporal models can branch16in the past and in the future, but we might want to consider a more17``modal'' conception of branching, in which we consider sequences of18events as possible histories. This does not necessarily require19changing our language, though we might also add our ``ordinary'' modal20operators $\Box$ and~$\Diamond$, and we could also consider adding21epistemic accessibility relations to represent changes in agents'22knowledge over time.2324\begin{defn}25 A \emph{possible histories model} for the temporal language is a triple26 $\mModel{M} = \tuple{T, C, V}$, where27 \begin{enumerate}28 \item $T$ is a nonempty set, interpreted as states in time.29 \item $C$ is a set of computational paths, or \emph{possible30 histories} of a system. In other words, $C$~is a set of31 sequences~$\sigma$ of states $s_1$, $s_2$,~$s_3$, \dots, where32 every $s_i \in T$.33 \item $V$ is a function assigning to each !!{propositional34 variable}~$p$ a set~$V(p)$ of points in time.35 \end{enumerate}36 To make things simpler, we will also generally assume that when a37 history is in~$C$, then so are all of its suffixes. For example, if38 $s_1$, $s_2$,~$s_3$ is a sequence in~$C$, then so are $s_2$, $s_3$39 and~$s_3$. Also, when two states $s_i$ and~$s_j$ appear in a40 sequence~$\sigma$, we say that $s_i \prec_\sigma s_j$ when $i < j$.41 When $t \in V(p)$ we say $p$~is \emph{true at}~$t$.42\end{defn}4344The one relevant change is that when we evaluate the truth of45!!a{formula} at a point in time~$t$ in a model~$\mModel M$, we do so46relative to a history~$\sigma$, in which $t$ appears as a state. We do47not need to change any of the semantics for !!{propositional48variable}s or for truth-functional connectives, though. All of those49are exactly as they were in \olref[sem]{defn:tmodels}, since none of50those will make reference to~$\sigma$. However, we now redefine our51future operator~$\Ftemp$ and add our~$\Diamond$ operator with respect52to these histories. 5354\begin{defn}\ollabel{defn:phmodels}55 \emph{Truth of !!a{formula}~$!A$ at~$t, \sigma$} in~$\mModel M$, in symbols:56 $\mSat{M}{!A}[t, \sigma]$:57 \begin{enumerate}58 \item\ollabel{defn:sub:phmodels-f} \indcase{!A}{\Ftemp59 !B}{$\mSat{M}{\indfrm}[t, \sigma]$ iff $\mSat{M}{!B}[t', \sigma]$60 for some $t' \in T$ such that $t \prec_\sigma t'$.}61 \item\ollabel{defn:sub:phmodels-diamond} \indcase{!A}{\Diamond62 !B}{$\mSat{M}{\indfrm}[t, \sigma]$ iff $\mSat{M}{!B}[t, \sigma']$63 for some $\sigma' \in C$ in which $t$ occurs.}64 \end{enumerate} 65\end{defn}6667Other temporal and modal operators can be defined similarly. However,68we can now represent claims that combine tense and modality. For69example, we might symbolize ``$p$~will not occur, but it might have70occurred'' using the formula $\lnot \Ftemp p \land \Diamond \Ftemp p$.71This would hold at a point and a history at which $p$ does not become72true at a successor state, but there is an alternative history at73which $p$ will become true. 7475\end{document}