See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/46666490

Logic in AI

Article · January 2006

Source: OAI

CITATIONS READS
0 7,601

1 author:

John-jules Meyer
Utrecht University
662 PUBLICATIONS 11,808 CITATIONS

SEE PROFILE

All content following this page was uploaded by John-jules Meyer on 20 June 2014.

The user has requested enhancement of the downloaded file.

ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

LOGIC IN AI

J.-J.Ch. Meyer
Department of Computing Science, Utrecht University, The Netherlands

Keywords: propositional logic, first-order predicate logic, second-order logic,

deduction theorem, (semi-)decidability, modal logic, possible world semantics,
intelligent agent, BDI (belief, desire, intention), dynamic logic, deontic logic, temporal
logic, epistemic logic, nonmonotonic logic, default logic, frame problem, qualification
problem, closed world assumption, circumscription, preferential semantics, multi-
valued logic, fuzzy logic.

Contents

S
TE S
1. Introduction
2. Classical Logic

R
AP LS
2.1 Classical prepositional logic
2.2 First-Order Predicate Logic
3. Modal Logic
C EO
3.1 Dynamic logic
3.2 Deontic logic
3.3 Temporal logic
3.4 Epistemic logic
E –

4. Nonmonotonic Logic
H
4.1 Default Logic
4.2 Circumscription
PL O

5. Multi-Valued Logic and Fuzzy Logic

6. Conclusion
M SC

Glossary
Bibliography
Biographical Sketch
SA NE

Summary

In this article we will give an overview of the role of logic in artificial intelligence.
U

After a general discussion of this role we turn to a more detailed treatment of classical
prepositional and first-order logic, and next to the most popular nonclassical logics in
AI, viz. modal logic, nonmonotonic logic and multi-valued logic.

1. Introduction

The discipline of artificial intelligence (AI) studies the question of how artifacts can be
ascribed or endowed with intelligence. In other words, AI concerns questions such as to
how to 'implement intelligence into artificial systems'. Of course, since the very concept
of intelligence is not fully understood, this means that in order to do something sensible,
one has to use some kind of working definition. In his (philosophical) introduction to AI
Copeland defines an agent (artificial or ortherwise) to be intelligent if it is 'massively
adaptable', i.e. if the agent is flexible to the degree that it can cope with all kinds of

©Encyclopedia of Life Support Systems (EOLSS)

ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

changes of situations in the world it's inhabiting. The latter may, of course, be the real
physical world we all live in, but one may also think of a more artificial world, such a.s
the internet or some computer-generated vitual reality.

Since intelligence often seems to involve some kind of reasoning it becomes clear that
logic, the science of reasoning, may play an important role in AI. This is true to the
extent that one adheres to the view that indeed intelligence has to do with or even can be
described in terms of symbolic means. At the moment there appear to be at least two
tendencies among AI researchers, one which holds that intelligence is to be described
and implemented in a symbolic way, and one which maintains that this view is
inadequate for implementing intelligence onto an artificial system. The former group of
reseachers ('the symbolists') indeed ascribe an important role to logic. Here the idea is to
lay down intelligent behaviour in formal/logical rules which can then be programmed
(by means of some programming language such as LISP or PROLOG) into a system.

S
TE S
The latter group does not believe that this can be done (since it is too complex, for
example), and claims that one has to resort to other ('non-symbolic' or 'subsymbolic')

R
AP LS
means like techniques inspired by biological organisms (like neural networks and
evolutionary computing mechanisms).
C EO
As to the symbolistic approach to AI one may again have different views as to the exact
role of logic in this enterprise. For example, is logic itself to be considered / employed
as a programming language, or at least as a kind of executable specification language, or
does logic 'merely' serve as an intermediary to get the concepts right and precise, after
which one may implement these by means of a procedural programming language?
E –

Also, there is the question which logic is to be employed. Contrary to what one may
H
think, there are many different logics. There is the familiar classical (prepositional and
PL O

first-order predicate) logic, but especially in the last century there have been developed
many 'non-classical' logics, which typically focus around some particular feature of
M SC

reasoning.

In this chapter we will briefly sketch some of these logics and indicate their role / use in
SA NE

AI. (More about the role of logic in AI and knowledge representation can be found in
the references.)

2. Classical Logic
U

The logic (or rather logics) usually referred to as classical logic comprises classical
prepositional and first-order logic. Although these logics may sometimes be regarded as
insufficient for AI purposes (as we will see below in subsequent sections), they
nevertheless have had a tremendous impact on later developments, and, moreover,
within AI these classical logics are still widely employed in many applications. For
example, many ways of knowledge representation in knowledge-based systems are still
using classical logic, and also the influential logic programming paradigm (such as the
AI language PROLOG) is primarily based upon classical logic. For this reason, as well
as for the reason that it provides the basis for many 'non-classical' logics', we here give a
succinct treatment of classical prepositional and first-order logic. (More can be found in
the many textbooks and handbooks on this subject, such as listed in the bibliography.)

©Encyclopedia of Life Support Systems (EOLSS)

ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

2.1 Classical prepositional logic

As indicated by its name prepositional logic is about reasoning about propositions, i.e.
assertions that can be true or false. The logic is therefore built on a set of primitive or
atomic propositions, sometimes also called atoms. Let's call this set℘. Now complex
propositions may be constructed from the primitive ones by using connectives, such as
'logical and' (∧), 'logical or' (∨), 'logical negation' (¬), 'logical implication' (→) and
'logical equivalence' or 'logical bi-implication' (↔). We denote formulas by ϕ,ψ,χ,
possibly endowed with marks and indexes. So complex formulas may have the
following form: ϕ∧ψ, ϕ∨ψ, ¬ϕ, ϕ→ψ, ϕ↔ψ.

The meaning of formulas in classical prepositional logic is given by assigning truth

values to these formulas on the basis of an assignment of truth values to the primitive
propositions. So let the valuation function π be a function that assigns truth values tt

S
TE S
(true) and ff (false) to the primitive atoms, i.e. π is a function ℘ → T, where T = {tt,ff}.
This means that the function π assigns a truth values for any atom p ∈ ℘: π(p) ∈ T.

R
AP LS
The prepositional meaning (i.e. truth value) V of a complex formula is now obtained by
considering the way such a formula is constructed. For convenience, we order the truth
C EO
values in T as follows: ff ≤ ff, ff ≤ tt, tt ≤ tt, and use functions min and max on truth
values, taking the minimum and maximum values, respectively, according to this
ordering. Furthermore, we have a function compl which takes the complement of a truth
value, i.e. compl(tt) = ff and compl(ff) = tt.
E –
H
ƒ V(p)=π(p) if p∈℘;
PL O

ƒ V(ϕ∧ψ)=min(V(ϕ),V(ψ));
ƒ V(ϕ∨ψ)=max(V(ϕ),V(ψ));
M SC

ƒ V(¬ϕ) = compl(V(ϕ));
ƒ V(ϕ→ψ) = max(compl(V(ϕ)), V(ψ));
ƒ V(ϕ↔ψ)=min(V(ϕ→ψ),V(ψ→ϕ))
SA NE

So, for example, if it is given that the atoms p is true and q is false (i.e. for p,q ∈℘ we
have that π(p) = tt and π(q) = ff), then we have that the formula p → q is false, since
V(p → q)= max(compl(V(p)),V(q)) = max(compl(π(p)), π (q)) = max(compl(tt),ff) =
U

max(ff, ff) = ff.

An alternative way of formulating the truth of a formula, which will easily generalize to
some of the other logics we will see in the sequel, is the following. We consider the
notion of a model that makes a formula true. Since in prepositional logic a model is just
a valuation function V, we will write V |= ϕ to denote that model V makes ϕ true.
Strictly speaking the valuation function V depends on the underlying truth assignment
function π, and should be written Vπ. However, we omit this superscript for notational
convenience.

Now we have that:

ƒ V |= p iff π (p) = tt for p∈℘;

©Encyclopedia of Life Support Systems (EOLSS)

ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

ƒ V |= ϕ∧ψ iff V |= ϕ and V |= ψ;

ƒ V|= ϕ∨ψ iff V |= ϕ or V |= ψ;
ƒ V |= ¬ϕ iff not V |= y ;
ƒ V |= ϕ→ψ iff V |= ϕ implies V |= ψ
ƒ V |= ϕ↔ψ iff V |=ϕ bi-implies V |= ψ.

Formulas ϕ that obtain the truth value tt no matter what the truth values of the atoms
occurring in it are (or, more formally, V(ϕ) = tt for all functions π assigning truth values
to the atoms), are called valid or tautological. If y is valid, we denote this by |= ϕ. These
formulas express 'necessary' truth under all circumstances. Examples of tautologies are:

ƒ |= ϕ→ϕ
ƒ |= (ϕ∧ψ) →ϕ
ƒ |= ϕ → (ϕ∨ψ)

S
TE S
ƒ |= ϕ ↔ ¬¬ϕ

R
AP LS
ƒ |= ϕ ∨ ¬ϕ
ƒ |= ¬(ϕ∧ψ)↔(¬ϕ∨¬ψ)
ƒ |= ¬(ϕ∨ψ)↔(¬ϕ∧¬ψ)
ƒ
C EO
|= ¬ϕ→(ϕ →ψ)
ƒ |= ϕ→(ψ∨¬ψ)

The first tautology should be obvious. The second one expresses a property of
E –

conjunction: a conjunction implies its conjuncts. The third one states a a property of
H
disjunction: one can always weaken a logical assertion by adding a disjunct. The fourth
one is the law of double negation: in classical logic double negation can be eliminated.
PL O

The fifth one is the well-known law of the excluded middle. The sixth and seventh ones
are the rules of De Morgan. The eighth and ninth ones are versions of the so-called
M SC

paradoxes of the material implication: they state that from a false assertion anything
follows, and that anything implies a true assertion, respectively.

Another important notion is that of (semantic) entailment: Φ |= ψ (where Φ is a set of

SA NE

formulas and ψ is a formula) expresses that if all formulas ϕ ∈ Φ are true with respect
to a valuation function, also the formula ψ is true with respect to that valuation function:
for any V it holds that if V(ϕ) = tt for all ϕ ∈ Φ, then also V(ψ) = tt. Another way of
U

saying the same is: "all models of the formulas ϕ ∈ Φ, are also models of the formula
ψ". (Recall that models are here the valuation functions V.) Examples are: ϕ |= ϕ and
ϕ∧ψ |= ϕ. There is an elegant relation between semantic entailment and validity, which
is called the (semantic version of the) deduction theorem:

Theorem 2.1 For any formulas ϕ1,…, ϕm it holds that:

ϕ1,…, ϕm |=ψ ⇔ |= (ϕ1 ∧…∧ ϕm) → ψ

In order to establish whether a formula is valid one may perform a systematic

enumeration of all possibilities for truth assignments for the atoms ("truth tables"), but
there are also other methods of answering this question, such as 'tableaux' or employing

©Encyclopedia of Life Support Systems (EOLSS)

ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

an axiomatic system. In the latter method one uses axioms and rules to try and derive
the formula at hand, starting with an axiom (or several of them) and then applying rules.
If one succeeds one knows that the formula is indeed a valid formula. For this to work it
is necessary to employ an axiom system that is so-called sound and complete.
Soundness of an axiomatic system means that every axiom is indeed valid and every
rule preserves validity, while completeness of an axiomatic system says that every valid
formula can be derived by means of the system (by using the axioms and rules) within
some finite number of steps (if one is clever enough!).

An example of a validity-preserving rule is the well-known modus ponens rule: if |= ϕ

and |= ϕ →ψ, then also |= ψ. This rule is often written as

ϕ, ϕ →ψ
ψ

S
TE S
There are many sound and complete axiom systems known for classical prepositional

R
AP LS
logic. A disadvantage of the axiomatic method is that if one does not succeed in
deriving the formula at hand, it is not clear whether the formula is indeed non-derivable
(and thus not valid by the completeness of the system) or it is the case that one has not
C EO
yet tried enough to obtain it. In fact, one can only really show that a formula ϕ is not
valid by giving a so-called countermodel, i.e. a truth assignment π for which V(ϕ) = ff.

2.2 First-Order Predicate Logic

E –
H
First-order predicate logic is a first extension of classical prepositional logic in the sense
PL O

that now the primitive propositions from prepositional logic get more structure. For
instance, in first-order predicate logic one may express that all objects in a room are red.
M SC

To this end the syntax of the logic is equipped with so-called predicate, constant and
function symbols. The predicate symbols are used to denote properties of elements in
the domain of discourse and relations between them. The constant symbols are used to
refer to elements in the domain. The function symbols are used to denote particular
SA NE

mathematical functions on the domain and thus may be used to refer indirectly to
particular objects in that domain as well.

Both predicate symbols and function symbols have a so-called arity, that is the number
U

of arguments they have. For instance, a predicate symbol '>' that will be used for
denoting the relation 'greater than' on a domain of numbers will typically have arity 2: >
(m1, m2), or mostly written as ' m1 > m2'. A predicate symbol R denoting a property like
'Red' will typicaly have arity 1: R(x). But in general we may have predicate symbols
with arity n. A function symbol '+' denoting the mathematical function of addition will
have arity 2: +( m1, m2), mostly written as m1 + m2, whereas a function symbol '√' for
the square root function will have arity 1: √x. By the way, constant symbols may be
regarded as function symbols with arity 0, while predicate symbols with arity 0 may be
viewed as prepositional primitives in the sense of prepositional logic!

We also have a set Var of (first-order) variables in our language which we will use in
(universally or existentially) quantified expressions. Terms are either variables,
constants or functions applied to terms (as many as the arity of the function). Atomic

©Encyclopedia of Life Support Systems (EOLSS)

ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

formulas are of the form P(t1, t2 ,..., tn), where P is a predicate symbol of arity n, and t1,
t2 ,..., tn are terms. Formulas are either atomic or formulas connected by the usual logical
connectives ∧, ∨, ¬, →, ↔, or by using the universal and existential quantifier: ∀xϕ
and ∃xϕ. (Here x is a variable, which typically (but not necessarily) will also occur in
the formula ϕ.) The reading of a formula of the form ∀xP(x) is "for all elements in the
domain it holds that they have property P", while ∃xP(x) reads "there exists an element
in the domain for which property P holds. "

We call an occurrence of a variable x free in a formula ϕ if that occurrence does not

occur in the scope of a quantifier, or put in other words, if that occurrence is not bound
by some quantifier. The models for an first-order language is a structure M = (D,I),
where D is a non-empty set, the domain of discourse, the objects which we talk about
(e.g. the natural numbers, but it can be anything), and I is an interpretation function that
interprets the predicate, constant and function symbols as relations, elements and

S
TE S
mathematical functions on the domain V, respectively. E.g. if D is the set of natural

R
numbers, the interpretation I(>) of a predicate symbol '>' may be the greater than

AP LS
relation on the natural numbers, the interpretation I(0) of a constant symbol '0' may be
the number zero, and the interpretation I(+) of the function symbol '+' may be the
mathematical function of addition.
C EO
To interpret variables we need a valuation function v of type Var → V, yielding for each
variable from the set Var a value in the domain V. In the sequel we will use the notation
v[d/x] for the modified valuation which is like v, but with x set to d. More formally,
E –
H
v[d/x](y) = d, if y = x;
PL O

v(y), otherwise

The interpretation of terms is now given inductively by a function IM,v as follows:

M SC

ƒ IM,v(c) = I(c) if c is a constant symbol;

ƒ IM,v (x) = v(c) if c is a variable in Var;
SA NE

ƒ IM,v (F) (t1 ,..., tn )= I(F)( IM,v (t1) ,..., IM,v (tn)) if F is a function symbol of arity n

Finally the truth of formulas given a model M and a valuation v, denoted M, v |= ϕ is

given inductively by:
U

ƒ M, v |= P(t1,... , tn) iff I(P)( IM,v (t1),..., IM,v (tn)) holds.

ƒ M, v |= ϕ ∧ψ iff M, v |= ϕ and M, v |= ψ
ƒ M, v |=ϕ ∨ψ iff M, v |= ϕ or M, v |= ψ
ƒ M, v |=¬ϕ iff not M, v |= ϕ
ƒ M, v |=ϕ →ψ iff M, v |=ϕ implies M, v |=ψ
ƒ M, v |=ϕ ↔ ψ iff M, v |=ϕ bi-implies M, v |=ψ
ƒ M, v |=∀xϕ iff M, v[d/x] |=ϕ for all d ∈ D;
ƒ M, v |=∃xϕ iff M, v[d/x] |=ϕ for some d ∈ D

A formula ϕ is called (universally) valid (denoted |= ϕ) iff it is true in every model and
for every valuation, i.e. M, v |=ϕ for any first-order model M and every valuation

©Encyclopedia of Life Support Systems (EOLSS)

ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

function v. A few examples of valid formulas are:

ƒ |= ∀xϕ ↔¬∃x¬ϕ
ƒ |= ∃xϕ ↔¬∀x¬ϕ
ƒ |= ∀x (ϕ ∧ψ ) ↔ (∀x ϕ ∧∀x ψ)
ƒ |= ∃x (ϕ ∨ψ ) ↔ (∃x ϕ ∨∃x ψ)
ƒ |= ∀x∀yϕ ↔∀y∀xϕ
ƒ |= ∃x∃yϕ ↔∃y∃xϕ
ƒ |= ∀x ϕ → ∃x ϕ
ƒ |= (∀x ϕ ∨∀x ψ )→ ∀x (ϕ ∨ ψ )
ƒ |=∃x (ϕ ∧ ψ ) → ( ∃x ϕ ∧∃x ψ )
ƒ |= ∃x∀y ϕ→ ∀y∃x ϕ
ƒ |= ∀x (ϕ→ψ)→ (∀x ϕ→∀x ψ)
|= ∀x (ϕ→ψ)→ (∃x ϕ→∃x ψ)

S
ƒ

TE S
ƒ |= ∀x (ϕ→ψ)↔ (ϕ→∀x ψ) if x does not occur free in the formula ϕ

R
AP LS
ƒ |= ∀x (ϕ→ψ)↔ ((∃x ϕ)→ ψ) if x does not occur free in the formula ψ

Again we have that modus ponens is a validity-preserving rule in the sense that if |= ϕ
C EO
and |= ϕ→ψ, then also |= ψ. Also we can define a notion of (semantic) entailment for
which the deduction theorem holds. We will not pursue this here. As for classical
prepositional logic there are also again sound and complete proof systems for first-order
predicate logic. However, there is not a mechanical method to decide the validity of a
formula as was the case in propositional logic: first-order predicate logic is only semi-
E –
H
decidable. If a formula is a validity one is able to derive it by means of a complete proof
system. However, while trying to derive a formula there is no way to tell whether this
PL O

procedure will ever end. Technically this means, although the validities of first-order
logic are recursively enumerable (by means of an algorithm, more or less deriving all
M SC

theorems of a complete proof system in a systematic way), the set of validities is not
decidable: a formula is not to be decided as a validity within a finite amount of time.
SA NE

Finally a remark on higher-order logic. It is possible to extend first-order logic with

high-order variables ranging over sets of domain elements (thus ranging over properties
and relations), but this results in such an expressive logic that there is not even a
complete proof system any more. So the validity in this logic is not even semi-decidable
U

any more. However, we'll see in the sequel that higher-order logic may come in handy
when we are interested in certain forms of non-monotonic logic, viz. circumscription.

-
-
-

TO ACCESS ALL THE 23 PAGES OF THIS CHAPTER,

Visit: http://www.eolss.net/Eolss-sampleAllChapter.aspx

©Encyclopedia of Life Support Systems (EOLSS)

ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

Bibliography

General note on the references. The references contain a number of (articles in) handbooks. These
generally give excellent surveys, but mostly they are written for an advanced audience. Also a number of
"Readings in" books are referenced. These are mostly collections of seminal papers in the field
(sometimes really defining or starting it), and are thus worthwhile for further reading.
Ph. Besnard, An Introduction to Default Logic, Springer, Berlin, 1989.[An excellent technical text book
for default reasoning]
R.J. Brachman & H.J. Levesque (eds.), Readings in Knowledge Representation, Morgan Kaufmann, Los
Altos, California, 1985.
G. Brewka, Nonmonotonic Reasoning: Logical Foundations of Common-sense, Cambridge University
Press, Cambridge, 1991.
F.M. Brown (ed.), The Frame Problem in Artificial Intelligence, Morgan Kaufmann, Los Altos CA, 1987.
B.F. Chellas. Modal Logic: An Introduction, Cambridge University Press, Cambridge/London 1980.

S
TE S
[Classical technical text book on modal logic]

R
AP LS
K. Clark, Negation as Failure, in: H. Gallaire & J. Minker (eds.), Logic and Data Bases, Plenum Press,
1978, pp. 293-322.
P.R. Cohen & H.J. Levesque, Intention is Choice with Commitment, Artificial Intelligence 42, 1990, pp.
213-261. [Seminal paper on the description of intelligent agents by means of modal (BDI-like) logics]
C EO
B..J. Copeland, Artificial Intelligence, A Philosophical Introduction, Black-well, Oxford, 1993. [An
excellent introduction to the philosophical issues in artificial intelligence]
R. Fagin, J.Y. Halpern, Y. Moses & M.Y. Vardi, Reasoning about Knowledge, The MIT Press,
Cambridge, Massachusetts, 1995. [Comprehensive textbook on epistemic logic and its application in AI
E –

and computer science]

H
D.M. Gabbay, I. Hodkinson & M. Reynolds, Temporal Logic, Mathematical Foundations and
PL O

Computational Aspects, Clarendon Press, Oxford. 1994.

D.M. Gabbay, C.J. Hogger &; J.A. Robinson (eds.), Handbook of Logic in Artificial Intelligence and
M SC

Logic Programming Vol. 1, Logical Foundations, Clarendon Press, Oxford, 1993.

D.M. Gabbay, C.J. Hogger & J.A. Robinson (eds.), Handbook of Logic in Artificial Intelligence and
Logic Programming Vol. 3, Nonmonotonic Reasoning and Uncertain Reasoning, Clarendon Press,
Oxford, 1994.
SA NE

D.M. Gabbay, C.J. Hogger & J.A. Robinson (eds.), Handbook of Logic in Artificial Intelligence and
Logic Programming Vol.4, Epistemic and Temporal Reasoning, Clarendon Press, Oxford, 1995.
M.R. Genesereth & N.J. Nilsson, Logical Foundations of Artificial Intelligence, Morgan Kaufmann, Los
U

Altos, 1987. [A very good textbook in which many logical subjects in relation to AI are discussed
succinctly and not too technically]
M.L. Ginsberg (ed.), Readings in Nonmonotonic Reasoning, Morgan Kaufmann, Los Altos, 1987.
R. Goldblatt, Logics of Time and Computation, CSLI Lecture Notes 7, Stanford, 1987. [An excellent
technical textbook on temporal and dynamic logic]
J.Y. Halpern & Y.O. Moses, Towards a Theory of Knowledge and Ignorance, in: Proceedings of the
AAAI Workshop on Non-Monotonic Reasoning, AAAI, 1984, pp. 125-143.
D. Harel, Dynamic Logic, in: D. Gabbay & F. Guenthner (eds.), Handbook of Philosophical Logic, Vol.
II, Reidel, Dordrecht/Boston, 1984, pp. 497-604.
G.E. Hughes & M.J. Cresswell, An Introduction to Modal logic, Methuen & Co. Ltd, London, 1968.
[Classical technical text book on modal logic]
D.J. Israel, The Role(s) of Logic in Artificial Intelligence, in: D.M. Gab-bay, C.J. Hogger & J.A.
Robinson (eds.), Handbook of Logic in Artificial Intelligence and Logic Programming Vol. 1, Logical

©Encyclopedia of Life Support Systems (EOLSS)

ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

Foundations, Clarendon Press, Oxford, 1993, pp. 1-29.

A. Kandel, Fuzzy Mathematical Techniques with Applications, Addison-Wesley, Reading MA, 1986.
S. Kraus, D. Lehmann & M. Magidor, Nonmonotonic Reasoning. Preferential Models and Cumulative
Logics, Artificial Intelligence 44, 1990, pp. 167-207. [A seminal paper on preferential semantics for
nonmonotonic logics]
S. Kripke, Semantic Analysis of Modal Logic, Zeitschnft fur Mathematische Logik und Grundlagen der
Mathematik 9, 1963, pp. 67-96. [A seminal paper on the possible world-style semantics for modal logic]
V. Lifschitz, Circumscription, in: D.M. Gabbay, C.J. Hogger & J.A. Robinson (eds.), Handbook of Logic
in Artificial Intelligence and Logic Programming Vol. 3. Nonmonotonic Reasoning and Uncertain
Reasoning, Clarendon Press, Oxford, 1994, pp. 297-352.
W. Lukaszewicz, Non-Monotonic Reasoning (Formalisation of Common-sense Reasoning), Ellis
Horwood, New York, 1990.
G. Malinowski, Many-Valued Logics, Clarendon Press, Oxford, 1993.

S
TE S
V. W. Marek &: M. Truszczynski, Nonmonotonic Logic, Context-Dependent Reasoning, Springer-Verlag,
Berlin, 1993. [A textbook on the mathematics behind nonmonotonic logics]

R
AP LS
J. McCarthy, Circumscription: a Form of Non-Monotonic Reasoning, Artificial Intelligence 13, 1980, pp.
27-39. [Seminal paper on the first forms of monotonic logic, viz. circumscription and default logic]
J.-J. Ch. Meyer & W. van der Hoek, Epistemic Logic for AI and Computer Science, Cambridge
C EO
University Press, Cambridge, 1995. [Comprehensive textbook on epistemic logic and its application in AI
and computer science]
J.-J. Ch. Meyer & W. van der Hoek, Modal Logics for Representing Incoherent Knowledge, in:
Handbook of Defeasible Reasoning and Uncertainty Management Systems, Vol. 3: Reasoning with Actual
and Potential Contradiction (Ph. Besnard & A. Hunter, eds.), Kluwer, Dordrecht/Boston, 1998, pp. 37-
E –

75.
H
J.-J. Ch. Meyer & R.J. Wieringa, Deontic Logic in Computer Science: Normative System Specification,
PL O

Wiley, 1993.
A. Ramsay, Formal Methods in Artificial Intelligence, Cambridge University Press, Cambridge, 1988.
M SC

A.S. Rao and M.P. Georgeff, Modeling Rational Agents within a BDI-Architecture, in: J. Alien, R. Fikes
& E. Sandewall (eds.), Proceedings of the Second International Conference on Principles of Knowledge
Representation and Reasoning (KR'91), Morgan Kaufmann, Los Altos, California, 1991, pp.473-484.
[Seminal paper on the description of intelligent agents by means of modal (BDI-like) logics]
SA NE

R. Reiter, On Closed World Data Bases, in: H. Gallaire & J. Minker (eds.), Logic and Data Bases,
Plenum Press, New York, 1978, pp. 55-76.
R. Reiter, A Logic for Default Reasoning, Artificial Intelligence 13, 1980, pp. 81-132. [Seminal paper on
U

the first forms of monotonic logic, viz. circumscription and default logic]
N. Rescher, Many-Valued Logic, McGraw-Hill, New York, 1969. [A comprehensive survey of many
proposals for multi-valued logic]
J.C. Shepherdson, Negation as Failure, Completion and Stratification, in: D.M. Gabbay, C.J. Hogger &
J.A. Robinson (eds.), Handbook of Logic in Artificial Intelligence and Logic Programming Vol.5, Logic
Programming, Clarendon Press, Oxford, 1998.
R. Turner, Logics for Artificial Intelligence, Ellis Horwood/Wiley. Chich-ester/New York, 1984. [A nice
little book which may spark your interest in nonclassical logics as applied in AI (it did so in my case)]
J. van Benthem, A Manual of Intensional Logic (2nd ed.), CSLI. Menio Park, 1988.
J. van Benthem, The Logic of Time (2nd ed.), Kluwer, Dordrecht/Boston, 1991.
W. van der Hoek, B. van Linder & J.-J. Ch. Meyer, An Integrated Modal Approach to Rational Agents,
in: M. Wooldridge & A. Rao (eds.), Foundations of Rational Agency, Kluwer, Dordrecht/Boston, 1999,
pp. 133-167. [A survey paper on the KARO logic, which is a logic for specifying intelligent agents based

©Encyclopedia of Life Support Systems (EOLSS)

 ARTIFICIAL INTELLIGENCE – Logic in AI - J.-J.Ch. Meyer

on dynamic and epistemic logic]

R.A. Wilson & F.C. Keil, The MIT Encyclopedia of the Cognitive Sciences, The MIT Press, Cambridge,
MA, 1999.
L.A. Zadeh, Fuzzy Logic and Approximate Reasoning, Synthese 30, 1975, pp. 407-428.

Biographical Sketch

Prof. Dr. John-Jules Ch. Meyer is head of the Intelligent Systems Group of the Institute of Information
and Computing Sciences at Utrecht University, and currently research director of this institute and vice-
dean of the Faculty of Mathematics and Computer Science at this university. He holds a MSc in
Mathematics from Leyden University, and a PhD in Computer Science from the Free University
Amsterdam on a thesis, entitled "Programming Calculi Based on Fixed Point Transformations", on the
semantics of programming languages.
From 1985 to 1993 he was successively an assistent, associate and full professor at the Computer Science
Department of the Free University Amsterdam, the last on a chair "Logic for Distributed Systems and

S
TE S
Artificial Intelligence". From 1989 to 1993 he was also a (part-time) full professor of Theoretical
Computer Science at Nijmegen University. Since 1993 he has been a (full) professor at Utrecht

R
AP LS
University. Furthermore, he is currently the scientific director of the Dutch Graduate School in
Information and Knowledge-based Systems (SIKS), chairman of the Dutch Association of Logic and
Philosophy of Science, member of the boards of the Dutch National Platform for Computer Science
Research and the Dutch Association of Theoretical Computer Science. His research interests include
C EO
artificial intelligence, agent technology, cognitive robotics, applied logic, and semantics of programming
languages.
He has published over 150 papers on these subjects in international journals and conference proceedings.
He is an editor of the Journal of Applied Non-Classical Logic(s), Data and Knowledge Engineering, and
the Journal of Intelligent Agents & Multi-Agent Systems. He has co-authored and co-edited several books
E –

on agent technology, deontic logic, epistemic logic and nonmonotonic reasoning.

H
He has been a member of over 100 (national and international) PhD committees in the area of computer
PL O

science and AI, at which he acted 19 times as a promotor.

M SC
SA NE
U

©Encyclopedia of Life Support Systems (EOLSS)

View publication stats