“M19N41” — 2008/11/5 — 17:14 — page 535 — #1 i
International Journal of Information and Management Sciences
Volume 19, Number 4, pp. 535-570, 2008
A Literature Review on Artiﬁcial Intelligence
S. A. Oke
University of Lagos
Research on artiﬁcial intelligence in the last two decades has greatly improved perfor-
mance of both manufacturing and service systems. Currently, there is a dire need for an
article that presents a holistic literature survey of worldwide, theoretical frameworks and
practical experiences in the ﬁeld of artiﬁcial intelligence. This paper reports the state-of-
the-art on artiﬁcial intelligence in an integrated, concise, and elegantly distilled manner to
show the experiences in the ﬁeld. In particular, this paper provides a broad review of recent
developments within the ﬁeld of artiﬁcial intelligence (AI) and its applications. The work is
targeted at new entrants to the artiﬁcial intelligence ﬁeld. It also reminds the experienced
researchers about some of the issue they have known.
Keywords: AI, Neural Network, Business Eﬃciency, Genetic Algorithms, Fuzzy Logic.
In the 21st century artiﬁcial intelligence (AI) has become an important area of re-
search in virtually all ﬁelds: engineering, science, education, medicine, business, account-
ing, ﬁnance, marketing, economics, stock market and law, among others (Halal (2003),
Masnikosa (1998), Metaxiotis et al. (2003), Raynor (2000), Stefanuk and Zhozhikashvili
(2002), Tay and Ho (1992) and Wongpinunwatana et al. (2000)). The ﬁeld of AI has
grown enormously to the extent that tracking proliferation of studies becomes a diﬃcult
task (Ambite and Knoblock (2001), Balazinski et al. (2002), Cristani (1999) and Goyache
(2003)). Apart from the application of AI to the ﬁelds mentioned above, studies have
been segregated into many areas with each of these springing up as individual ﬁelds of
knowledge (Eiter et al. (2003), Finkelstein et al. (2003), Grunwald and Halpern (2003),
Guestrin et al. (2003), Lin (2003), Stone et al. (2003) and Wilkins et al. (2003)).
Received January 2007; Revised and Accepted September 2007.
“M19N41” — 2008/11/5 — 17:14 — page 536 — #2 i
536 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
1.1. The challenge of the AI ﬁeld
This work grew out of the challenges that AI possesses in view of the rise and grow-
ing nature of information technology worldwide that has characterised business- and
non-business organisational development (Barzilay et al. (2002), Baxter et al. (2001),
Darwiche and Marquis (2002), Gao and Culberson (2002), Tennenholtz (2002) and
The necessity for research in AI is being motivated by two factors that are (i) to
give the new entrants into the AI ﬁeld an understanding of the basic structure of the
AI literature (Brooks (2001), Gamberger and Lavrac (2002), Kim (1995), Kim and Kim
(1995), Patel-Schneider and Sebastiani (2003) and Zanuttini (2003)). As such, the litera-
ture discussed here answers the common query, “why must I study AI?” (ii) the upsurge
of interest in AI that has prompted an increased interest and huge investments in AI
Interested researchers from all disciplines wish to be aware of the work of others
in their ﬁeld, and share the knowledge gleaned over the years (Rosati (1999), Kaminka
et al. (2002), Bod (2002), Acid and De Campos (2003), Walsh and Wellman (2003),
Kambhampati (2000) and Barber (2000)). By sharing AI knowledge, new techniques
and approaches can be developed so that a greater understanding of the ﬁeld can be
gained. To these ends, this paper has also been written for researchers in AI so they
can continue in their eﬀorts aimed at developing this area of concentration through
newly generated ideas. Consequently, they would be able to push forward the frontier of
knowledge in AI.
In the section that follows this paper presents a brief explanation of some important
areas in Artiﬁcial Intelligence. This is to introduce the readers into the wide-ranging
topics that AI encompasses. In another section, a comprehensive review of the literature
along the major categories of artiﬁcial intelligence is presented. The review raises some
important questions with serious research implications for those who are interested in
carrying out research artiﬁcial intelligence. These questions if well addressed will solve
some unresolved technical and non-technical issues carried over from the last decade to
the present time.
1.2. An overview of the AI ﬁeld
On a very broad account the areas of artiﬁcial intelligence are classiﬁed into sixteen
categories (Becker et al. (2000), Singer et al. (2000), Chen and Van Beek (2001), Hong
“M19N41” — 2008/11/5 — 17:14 — page 537 — #3 i
A Literature Review on Artiﬁcial Intelligence 537
(2001) and Stone et al. (2001)). These are: reasoning, programming, artiﬁcial life, be-
lief revision, data mining, distributed AI, expert systems, genetic algorithms, systems,
knowledge representation, machine learning, natural language understanding, neural net-
works, theorem proving, constraint satisfaction, and theory of computation (Peng and
Zhang (2007), Zhou et al. (2007) and Wang et al. (2007)). Since many readers of this
article may require a glance view of the AI ﬁeld, the author has utilised a ﬂow diagram to
illustrate the whole structure of this paper, and the relationship among the diverse ﬁelds
of AI, as presented in Figure 1. What follows is a brief discussion of some of the impor-
tant areas of AI (Chan and Darwiche (2002), Pool and Zhang (2003), Bhattacharyya and
Keerthi (2001), Chawla et al. (2002), Al-Ani and Deriche (2002) and Xu and Li (2000)).
These descriptions only account for a selected number of areas.
The ﬁrst major area considered here is that of reasoning. Research on reasoning has
evolved from the following dimensions: case-based, non-monotonic, model, qualitative,
automated, spatial, temporal and common sense.
For an illustrative example, the case-based reasoning (CBR) is brieﬂy discussed. In
CBR, a set of cases stored in a case base is the primary source of knowledge. Cases
represent speciﬁc experience in a problem-solving domain, rather than general rules.
The main activities when solving problems with cases are described in the case-based
reasoning cycle. This cycle proposes the four steps: relieve, reuse, revise and retain.
First, the new problem to be solved must be formally described as a case (new case).
Then, a case that is similar to the current problem is retrieved from the case base.
The solution contained in this retrieved case is reused to solve the new problem with
a new solution obtained and presented to the user who can verify and possibly revise
the solution. The revised case (or the experience gained during the case-based problem
solving process) is then retained for future problem solving. Detailed information on
“dimensions” or how they are related could be obtained from the relevant sources listed
in the references (Debruyne and Bessiere (2001), Halpern (2000), Halpern (2001), Renz
and Nebel (2001), Singh et al. (2002) and Straccia (2001)).
1.2.2. Genetic algorithm
The second major area of AI treated here is Genetic Algorithm (GA). This is a
search algorithm based on the mechanics of natural selection and natural genetics. It is
“M19N41” — 2008/11/5 — 17:14 — page 538 — #4 i
538 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
Figure 1. Illustration concerning the relationship among the diverse ﬁelds of AI.
“M19N41” — 2008/11/5 — 17:14 — page 539 — #5 i
A Literature Review on Artiﬁcial Intelligence 539
an iterative procedure maintaining a population of structures that are candidate solu-
tions to speciﬁc domain challenges. During each generation the structures in the current
population are rated for their eﬀectiveness as solutions, an don the basis of these evalua-
tions, a new population of candidate structures is formed using speciﬁc genetic operators
such as reproduction, cross over and mutation.
1.2.3. Expert system
The third aspect of AI discussed here is expert system. An expert system is com-
puter software that can solve a narrowly deﬁned set of problems using information and
reasoning techniques normally associated with a human expert. It could also be viewed
as a computer system that performs at or near the level of a human expert in a particular
ﬁeld of endeavour.
1.2.4. Natural language understanding
Natural language generation (NLG) systems are computer software systems that
produce texts in English and other human languages, often from non-linguistic input
data. NLG systems, like most AI systems, need substantial amounts of knowledge that
is diﬃcult to acquire. In general terms, these problems were due to the complexity,
novelty, and poorly understood nature of the tasks the systems attempted, and were
worsened by the fact that people write so diﬀerently (Reiter et al. (2003)).
1.2.5. Knowledge representation (KR)
Knowledge bases are used to model application domains and to facilitate access to
stored information. Research on KR originally concentrated around formalisms that
are typically tuned to deal with relatively small knowledge base, but provide powerful
reasoning services, and are highly expressive.
2. The Artiﬁcial Intelligence Literature
2.1. Reasoning in artiﬁcial intelligence
The theory and practice of reasoning in artiﬁcial intelligence has extensive docu-
mentation (Atkinson and Bench-Capon (2007)). Researchers have worked in terms of:
(i) development of axioms that give sound and complete axiomazation for the logic of
reasoning; (ii) the theoretical properties of the algorithms used for qualitative temporal
“M19N41” — 2008/11/5 — 17:14 — page 540 — #6 i
540 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
reasoning; (iii) what is relevant to a given problem of reasoning (independence); (iv) and
methods for qualitative reasoning. A study on axomatising causal reasoning is credited
to Halpern (2000). The author axomatised causal models deﬁned in terms of a collection
of equations as deﬁned by Pearl.
Axiomatisations are provided for three successively more general classes of causal
models (i) the class of recursive theories (those without feedback); (ii) the class of the-
ories where the solutions to the equations are unique; (iii) arbitrary theories (where the
equations may not have solutions and, if they do, they are not necessarily unique). It is
shown that to reason about causality in the most general third class, we must extend the
language used by Galles and Pearl. In addition, the complexity of the decision procedures
is characterised for all the languages and classes of models considered.
The concept of reasoning in Artiﬁcial Intelligence has been discussed under some
general areas, which include complexity of reasoning, reasoning about minimal belief,
axiomatising, sampling algorithm, conditional plausibility, eﬃcient methods, logic and
consistency, fuzzy description logics, backbone fragility, diagnosis, independence, domain
ﬁltering, and fusion. The literature on complexity of reasoning relates to spatial con-
gruence and expressive description logics. Cristani (1999) introduces a novel algebra for
reasoning about spatial congruence, thus, showing that the satisﬁability problem in the
spatial algebra MC-4 is NP-complete, and present a complete classiﬁcation of tractability
in the algebra, based on the individuation of three maximal tractable sub classes, one
containing the basic relations.
The work by Tobies (2000) studies the complexity of the combination of the de-
scription logics ALCQ and ALCQI with a terminological formalism based on cardinality
restrictions on concepts. These combination can naturally be embedded into C∧2, the
two variable fragment of predicate logic with counting quantiﬁers, which yields decid-
ability in next time.
In another work, Cheng and Druzdzel (2000) develop an algorithm for evidential
reasoning in large Bayesian networks. An adaptive importance sampling algorithm, AIS-
BN that shows promising convergence rates even under extreme conditions is developed.
It seems to outperform the existing sampling algorithm consistently. This provides a
better substitute to stochastic sampling algorithms that have been observed to perform
poorly in evidential reasoning with extremely unlikely evidence.
The concept of conditional plausibility is well treated in Halpern (2001). Halpern
deﬁnes a general notion of algebraic conditional plausibility measures. It is shown that
“M19N41” — 2008/11/5 — 17:14 — page 541 — #7 i
A Literature Review on Artiﬁcial Intelligence 541
algebraic conditional plausibility measures can be represented using Bayesian networks.
On the issue of eﬃciency methods, Renz and Nebel (2001) analyse the theoretical
properties of qualitative spatial reasoning in the RCC8 framework. They demonstrate
that the orthogonal combination of heuristic methods is successful in solving almost all
apparently hard instances in the phase transition region up to a certain size in reasonable
In a paper, Rosati (1999) conceptualise the minimal belief and negation as fail-
ure (MBNF) in its prepositional fragment as introduced by Lifschitz. The concept can
be considered as a unifying framework for several non-monotonic formalisms, including
default logic, autoepistemic logic, circumstription, epistemic queries and logic program-
ming. The application of soft computing theory is vast in the reasoning literature. One
of such studies was carried out by Straccia (2001) on reasoning within fuzzy description
logics. The paper presents a fuzzy extension of ALC, combining Zadeh’s fuzzy logic with
a classical DL. The work supports the idea of managing structured knowledge with appro-
priate syntax, semantics and properties on constraint propagation calculus for reasoning
Singer et al. (2000) introduce the backbone fragility and the local search cost peak.
The authors introduce a temporal model for reasoning on disjunctive metric constraints
on intervals and time points in temporal contexts. This temporal model is composed of
a labeled temporal algebra and its reasoning algorithms. The computational cost of rea-
soning algorithms is exponential in accordance with the underlying problem complexity,
although some improvements were proposed.
On diagnosis, Console et al. (2003) extend the approach to deal with temporal in-
formation. They introduce a notion of temporal decision tree, which is designed to make
use of relevant information as long as it is acquired, and they present an algorithm for
compiling such trees from a model-based reasoning system. A noteworthy study that
considers independence was embarked upon by Lang et al. (2003). Two basic forms of
independence, namely, a syntactic one and a semantic one are treated. They also con-
sider the problem of forgetting, i.e. distilling from a knowledge base only the part that
is relevant to the set of queries constructed from a subset of the alphabet.
Still on the reasoning literature, Debruyne and Bessiere (2001) focuses on the local
consistencies that are stronger than arc consistency, without changing the structure of
the network, i.e., only removing inconsistent values from the domains. They compared
“M19N41” — 2008/11/5 — 17:14 — page 542 — #8 i
542 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
them both theoretically and experimentally, considering their pruning eﬃciency and the
time required enforcing them.
The fusion concept was treated in Baader et al. (2002). The authors extend the
decidability transfer results from normal modal logics to a large class of description
logics. They introduce abstract description systems, to cover diﬀerent description logics
in a uniform way which can be seen as a common generalisation of description and modal
logics, and show the transfer result in this general setting.
On the concept of logic in reasoning, Halpern and Pucella (2002) presents a preposi-
tional logic to reason about the uncertainty of events, where the uncertainty is modeled
by a set of probability measures assigning an interval of probability to each event. They
give a sound and complete axiomatisation for the logic, and show that the satisﬁability
problem is NP-complete, no harder than satisﬁability for prepositional logic.
An important research area in reasoning is on consistency. Wray and Laird (2003)
show how the combination of a hierarchy and persistent assertions of knowledge can lead
to diﬃculty in maintaining logical consistency in asserted knowledge. They explore the
problematic consequence of persistent assumptions in the reasoning process and introduce
novel potential solutions.
On constraint reasoning, Younes and Simmons (2003) present an adaptation of the
additive heuristic for plan space planning, and modify it to account for possible reuse of
existing actions in a plan. They also propose a large set of novel ﬂaw selection strategies,
and show how these can help them solve more problems than previously possible by
POCL planners. VHPOP also supports planning with durative actions by incorporating
standard techniques for temporal constraint reasoning.
2.2. Natural language understanding
The natural language literature broadly consists of many aspects. Within the limits
of the work reported here, scholars have investigated into the semantic, mapping, knowl-
edge acquisition and selection procedure of natural languages. The ﬁrst two listings deal
with representation of natural languages in a taxonomy form and linking the semantic
together in a group. Knowledge acquisition and selection have been treated from the
viewpoint of nature of tasks and informativeness of the problem considered.
In an article on semantic similarity in taxonomy, Resnik (1999) presents a measure of
semantic similarity in IS-A taxonomy based on the notion of shared information context.
The author presents algorithms that take advantage of taxonomic similarity in resolving
“M19N41” — 2008/11/5 — 17:14 — page 543 — #9 i
A Literature Review on Artiﬁcial Intelligence 543
syntactic and semantic ambiguity, along with experimental results demonstrating their
eﬀectiveness. The work gives a clear understanding of the concept of semantics, thus
improving the problem-solving viewpoint and conceptualisation of work in semantic.
In the paper by Thompson et al. (2003), the authors focus on a system, WOLFIE
(Word Learning from Interpreted Examples), that acquires a semantic lexicon from a
corpus of sentences paired with semantic representations. The work is useful in the
ability of the software developed to aid supervised learning since WOLFIE has ability to
learn useful lexicons for a database interface in four diﬀerent natural languages.
In another paper, Reiter et al. (2003), the idea of acquiring correct knowledge for
natural language generation was discussed. The authors identiﬁed a number of problems
that relates to knowledge acquisition such as complexity, novelty and poor understanding
nature of tasks. The problem could be worsened by the fact that people write so diﬀer-
ently. Thus, the authors have contributed through discussions on practical experiences.
The concept of committee-based sample selection for probabilistic classiﬁers was
discussed in Argamon-Engelson and Dagans (1999). The paper investigates methods for
reducing annotation cost by “sample selection”. The contribution of the authors hinges
on the fact that redundancy in labeling examples that contribute little new information
2.3. Genetic algorithm literature
Genetic algorithm is an important and growing part of the artiﬁcial intelligence
literature with numerous research ﬁndings. A good example of such studies could be
found in Turney (1995). The study introduces ICET, a new algorithm for cost-sensitivity
classiﬁcation. ICET uses a genetic algorithm to evolve a population of biases for a
decision tree induction algorithm. ICET is compared here with three other algorithms
for cost-sensitive classiﬁcation - EG2, CS-ID3, and IDX- and also with C4.5, which
classiﬁes without regard to cost.
2.4. Knowledge representation research
Knowledge representation is an important aspect of artiﬁcial intelligence research
with many dimensions (2003). The following is a cross section of studies carried out
on knowledge representation. In a study by Cadoli et al. (2000), the space eﬃciency
of prepositional knowledge representation (PKR) formalism was investigated. It is as-
sumed that knowledge is either a set of prepositional interpretations (models) or a set
“M19N41” — 2008/11/5 — 17:14 — page 544 — #10 i
544 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
of prepositional formulae (theorems). A formal way of talking about the relative ability
of PKR formalisms to compactly represent a set of models or a set of theorems was
provided. One interesting result is that formalisms with the same time complexity do
not necessarily belong to the same space eﬃciency class.
Yet in another work, Di-Sciascio et al. (2002) propose a structured approach to
the problem of retrieval of images by content and present a description logic that has
been devised for the semantic indexing and retrieval of images containing complex ob-
jects. Using the logical approach as a formal speciﬁcation, they implemented a complete
client-server image retrieval system, which allows a user to pose both queries by sketch
and queries by example. Results were presented adopting a well-established measure of
quality borrowed from textual information retrieval.
Kusters and Borgida (2001) studies the functional relationships between objects.
The authors show that although determining subsumption between concept descriptions
has the same complexity (through requiring diﬀerent algorithms), the story is diﬀerent
in the case of determining the least common subsumer (LCS).
In another study, Baget and Mugnier (2002) consider simple conceptual graphs as
the kernel of most knowledge representation formalisms built upon Sowa’s model. They
present a family of extensions of this model, based on rules and constraints, keeping
graph homomorphism as the basic operation. These results extend and complete the
ones already published by the authors.
In an interesting study, the notion of class representation formalism was investigated
(Calvanese et al., 1999). The basic issues underlying such representation formalisms
and single out both their common characteristics and their distinguishing features were
studied. The formalism is expressed in the style of description logic, which have been
introduced in knowledge representation as a means to provide a semantically well founded
basis for the structural aspects of knowledge representation systems.
2.5. Machine learning literature
The literature of machine learning is wide (Grumberg et al., 2003, Brodley and Friedl,
1999, Meek, 2001 and Walker, 2000).The following is a brief description of the various
machine learning articles. The paper by Schlimmer and Hermens (1993), describes an
interactive note-taking system for pen-based computers with two distinctive features.
The system is an example of a learning apprentice software-agent. A machine learning
component characterises the syntax and semantics of the users information.
“M19N41” — 2008/11/5 — 17:14 — page 545 — #11 i
A Literature Review on Artiﬁcial Intelligence 545
In another work, Soderland and Lehnert (1994) present a novel approach that uses
machine learning to acquire knowledge for some of the higher level IE processing. It was
found that performance equals that of a partially trainable discourse module requiring
manual customisation for each domain.
In Baxter (2000), a model of inductive bias learning was developed. The central
assumption of the model is that the learner is embedded within an environment of re-
lated learning tasks. Explicit bounds were also derived demonstrating that learning
multiple tasks within an environment of related tasks could potentially give much better
generalisation than learning a single task.
Blockeel et al. (2002) improve the eﬃciency of inductive logic programming through
the use of query packs. A complexity analysis by the authors shows that considerable
eﬃciency improvements can be achieved through the use of this query pack execution
mechanism. This claim is supported by empirical results obtained by incorporating
support for query pack execution in two existing learning systems.
Nock (2002) presents theoretical results, approximation algorithms, and experiments
on inducing interpretable voting classiﬁers without trading accuracy for simplicity. It
is ﬁrst attempt to build a voting classiﬁer as a base formula using the weak learning
framework (the one which was previously highly successful for decision tree induction),
and not the strong learning framework (as usual for such classiﬁers with boosting-like
approaches). Experimental results on thirty-one domains tend to display the ability of
WIDC to produce small, accurate and interpretable decision committees.
The machine learning literature also beneﬁted from the study due to Lerman et
al. (2003). The paper considers the Wrapper maintenance problem using a machine
learning approach. The authors present an eﬃcient algorithm that learns structural
information about data from positive examples alone. The Wrapper veriﬁcation system
detects when a wrapper is not extracting correct data, usually because the Web source
has changed its format. They were able to successfully reinduce the wrappers, obtaining
precision and recall values of 0.90 and 0.80 on the data extraction task.
Wolpert and Tumer (2002) consider the problem of designing the utility functions of
the utility-maximising agents in a multi-agent system so that they work synergistically
to maximise a global utility. The particular problem domain they explore is the control
of network routing by placing agents on all the routers in the network. They present
experiments verifying this, and also showing that a machine-learning-based version of
“M19N41” — 2008/11/5 — 17:14 — page 546 — #12 i
546 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
the COIN algorithm in which costs are only imprecisely estimated via empirical means
also outperforms the ISPA, despite having access to less information than the ISPA.
Still on machine learning, the important study due to Gordon (2000) should be
noted. The study develops agents that are adaptive and predictable and timely. The
paper is to improve the eﬃciency of re-veriﬁcation after learning, so that agent has a
suﬃciently rapid response time. The study presents two solutions: positive results that
certain learning operators are apriori guaranteed to preserve useful classes of behavioural
assurance constraints (which implies that no re-veriﬁcation is needed for these operators),
and eﬃcient incremental re-veriﬁcation algorithms for those learning operators that have
negative apriori results.
Still on the machine learning literature, Dietterich (2000) presents a new approach
to hierarchical reinforcement learning based on decamping the target Markov decision
process (MDP) into a hierarchy of smaller MDPs and decomposing the value function of
the target MDP into an additive combination of the value function of the smaller MDPs.
It demonstrates the eﬀectiveness of his non-hierarchical execution experimentally and
concludes with a comparison to related work and a discussion of the design tradeoﬀs in
hierarchical reinforcement learning.
The paper by Elomma and Kaariainen (2001) presents analysis of reduced error
pruning in three diﬀerent settings. It clariﬁes the diﬀerent variants of the reduced error
pruning algorithm, brings new insight to its algorithmic properties, analyses the algo-
rithm with less imposed assumptions than before, and includes the previously overlooked
empty subtrees to the analysis.
In another contribution to knowledge, GPOMDP, a simulation-based algorithm for
generating a biased estimate of the gradient of the average reward in partially observable
Markov decision process POMDPs controlled by parameterised stochastic policies was
introduced (Baxter and Bartlett, 2001). The authors prove the convergence of GPOMDP,
and show how the correct choice of the parameter beta is related to the mixing time
of the controlled POMDP and describe extensions of GPOMDP to controlled Markov
multiple agents, higher-order derivatives, and a version for training stochastic policies
with internal states.
In another work, Fern et al. (2002) develop, analyse and evaluate a novel, supervised,
speciﬁc-to-general learner for a simple temporal logic and use the resulting algorithm to
learn visual event deﬁnitions from video sequences. They apply the algorithm developed
“M19N41” — 2008/11/5 — 17:14 — page 547 — #13 i
A Literature Review on Artiﬁcial Intelligence 547
to the task of learning relational event deﬁnitions from video and show that it yields
deﬁnitions that are competitive with hand-coded ones.
Another interesting documentation could be credited to Brafma and Tennenholtz
(2003) who proposed a number of reinforcement learning algorithms and showed that
some converged to good solutions in the limit. They show that using very simple model-
based algorithms, much better convergence rates can be attained.
The work by Price et al. (2003) is another addition to the machine learning literature.
The authors propose and study a formal model of implicit imitation that can accelerate
reinforcement learning dramatically in certain cases. Though they make some stringent
assumptions regarding observability and possible interactions, they brieﬂy comment on
extensions of the model that relaxes these restrictions.
Xu et al. (2002) carried out a study on the recursive least-square (RLS) algorithm.
In the study, RLS methods are used to solve reinforcement learning problems, where
two new reinforcement learning algorithms using linear value function approximators
are proposed and analysed. The performance of fast AHC is also compared with that
of the AHC method using LS-TD (lambda). The experimental results were analysed
based on the existing theoretical work on the transient phase of forgetting factor RLS
methods. The study by Weiss and Provost (2003) is particularly focused on training
and development situations. Their study considered the learning situation when training
data are costly. The eﬀect of class distribution on tree induction was emphasised. The
article helps to answer the question, “if only n training examples can be selected, and in
what proportion should be classes be represented?”
By analysing, for a ﬁxed training set size, the relationship between the class distribu-
tion of training data and the performance of classiﬁcation trees induced from these data
is possible. An empirical analysis of this algorithm shows that the class distribution of
the resulting training set yields classiﬁers with good (nearly-optimal) classiﬁcation per-
formance. Drummond (2002) in a conceptual study discusses a system that accelerates
reinforcement learning by using transfer from related tasks. Experiments demonstrate
that function composition often produces more than an order of magnitude increase in
learning rate compared to a basic reinforcement learning algorithm.
In another work, a reinforcement learning approach for automatically optimising a
dialogue policy, which address the technical challenges in applying reinforcement learn-
ing to a working dialogue system with human users was presented (Singh et al., 2002).
“M19N41” — 2008/11/5 — 17:14 — page 548 — #14 i
548 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
The work reports on the design, construction and empirical evaluation of NJFun, an ex-
perimental spoken dialogue system that provides users with access to information about
fun things to do in New Jersey. The results show that by optimising its performance via
reinforcement learning, NJFun measurably improves system performance.
2.6. Theorem proving
In a very interesting study, the assumptions needed to prove Cox’s theorem are
discussed and examined by Halpern (1999). The various sets of assumptions under
which Cox-style theorem can be proved are provided, although all are rather strong and
arguably not natural.
2.7. Theory of computation
The paper by Ginsberg (2001) investigates the problems arising in the construction
of a program to play the game of contract bridge. GIB, the program being described,
involves ﬁve separate technical advances: partition search, the practical application of
Monte Carlo techniques to realistic problems, a focus on achievable sets to solve problems
inherent in the Monte Carlo approach, an extension of alpha-beta pruning from total
orders to arbitrary distributive lattices, and the use of squeaky wheel optimisation to ﬁnd
approximately optimal solutions to card-play problems. In another paper, an algorithm
for identifying noun-phrase antecedents of pronouns and adjectival anaphors in Spanish
dialogues was presented ( Palomar and Martines-Barco, 2001). The algorithm is based
on linguistic constraints and preferences and uses an anaphoric accessibility space within
which the algorithm ﬁnds the noun phrase. The algorithm is implemented in prolog.
According to this study, 95.9% of antecedents were located in the proposed space, a
precision of 81.3% was obtained for pronominal anaphora resolution, and 81.5% for
Koehler and Hoﬀmann (2000) addresses the problem of computing goal orderings,
which is one of the longstanding issues in AI planning. It makes two new contributions:
the paper formally deﬁnes and discusses two diﬀerent goal orderings; and developed
two diﬀerent methods to compute reasonable goal orderings. The complexity of these
orderings is investigated and their practical relevance is discussed. In a research on
non-approaximability results for partially obvservable Markov decision processes, Lusena
et al. (2001) show that for several variations of partially observable Markov decision
processes, polynomial-time algorithms for ﬁnding control policies are unlikely to or simply
“M19N41” — 2008/11/5 — 17:14 — page 549 — #15 i
A Literature Review on Artiﬁcial Intelligence 549
don’t have guarantees of ﬁnding policies within a constant factor or a constant summand
To provide a tool for use by multiagent researchers in evaluating this trade oﬀ, a
uniﬁed framework, the communicative multiagent team decision problem (COM-MTDP)
was presented (Pynadath and Tambe, 2002). The COM-MTDP model combines and ex-
tends existing multiagent theories, such as decentralised partially observable Markov
decision processes and economic team theory. In addition to their generality of rep-
resentation, COM-MTDPs also support the analysis of both the optimality of team
performance and the computational complexity of the agents’ decision problem.
Cemgil and Kappen (2003) present a probabilistic generative model for timing devi-
ations in expressive music performance. The structure of the proposed model is equiva-
lent to a switching state space model. They formulate two well known music recognition
problems, namely tempotracking and automatic transcription (rhythm quantisation) as
ﬁltering and maximum a posteriori (MAP) state estimation tasks. The simulation results
suggest better results with sequential methods. In the article on Bound Propagation,
Leisink and Kappen (2003) present an algorithm to compute bounds on the marginals of
a graphical model. This can be considered as a set of constraints in a linear programming
problem of which the objective function is the marginal probability of the center nodes.
They show that sharp bounds can be obtained for indirect and directed graphs that are
used for practical applications, but for which exact computations are infeasible.
Based on the Davis-Putnam procedure, Birnbaum and Lozinskii (1999) present an
algorithm, CDP, that computes the exact number of models of a prepositional CNF and
DNF formula F. The practical performance of CDP has been estimated in a series of
experiments on a wide variety of CNF formulae.
2.8. The programming literature
There are a large number of articles on programming in artiﬁcial intelligence. Since
programming is empirical based, most of the papers have sprang up for modeling or
mathematical frameworks. An example of an article in this domain is written by Sato and
Kameya (2001). The work hinges on parameter learning of logic programmes for symbolic
- statistical modeling. The authors deﬁned clause programs containing probabilistic
facts with a parameterised distribution. The work extent the traditional least Her brand
model semantics in logic programming in distribution semantics, possible world semantic
“M19N41” — 2008/11/5 — 17:14 — page 550 — #16 i
550 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
with a probability distribution which is unconditionally applicable to arbitrary logic
programmes including HMMs, PCFGs and Bayesian networks.
2.9. Neural networks research
On neural networks, Opitz and Machin (1999) carried out an empirical study of
the population ensemble methods. An ensemble consists of a set of individually trained
classiﬁes (such as neural networks or decision trees) whose predictions are combined when
classifying novel instances. Previous research has shown that an ensembles is often more
accurate than any of the single classiﬁers in the ensemble. The result clearly indicate
that bagging is sometimes much less accurate than boosting and that boosting can create
ensembles that are less accurate than a single classiﬁers. Further result shows that
boosting ensemble may often over ﬁt noisy data sets, thus decreasing its performance.
3. Applications of Artiﬁcial Intelligence
Studies on applications of AI are diverse (Andrew, 2001, Basu et al., 2001, Bui et al.,
2002, Peral and Ferrandez, 2003, Plenert, 1994 and Scerri et al., 2002). In the following
sub-sections, we present application-based studies.
3.1. Applications of AI in planning and scheduling
In recent years, research in the planning community has experienced a wide variety
of studies (Boutilier et al., 1999, Brafman and Domshlak, 2003, Cimatti and Roveri,
2000, Hauskrecht, 2000 and Howe and Dahlman, 2002). Research is increasingly moving
towards application of planners to realistic problems involving both time and many
types of resources. Some of the several planners developed include PDDL2.1, SHOP 2,
CRAPU PLAN, NADL, POMP, GRT, FF, PBR, TALplanner, AltAltp, MIPS, Metric-
FF Planning System, and SAPA (Refanidis and Vlahavas, 2001, Hoﬀman and Nebel,
2001, Kvarnstrom and Magnusson, 2003, Sanchez and Kambhampati, 2003, Hoﬀman,
2003 and Edelkamp, 2003). A brief account of research directions in planning is given
On planning and scheduling the paper by Long and Fox (2003) is very signiﬁcant
to the literature. The authors reported that interest in planning demonstrated by the
manufacturing research community has inspired work in observation scheduling, logistics
“M19N41” — 2008/11/5 — 17:14 — page 551 — #17 i
A Literature Review on Artiﬁcial Intelligence 551
planning, and plant control. Intensive eﬀorts have also been made to focus the commu-
nity on the modelling and reasoning issues that must be confronted to make planning
technology meet the challenges of application. Long and Fox (2003) reasoned that the
international planning competitions have acted as an important motivating force behind
the progress that has been made in planning since 1998. The third competition (held in
2002) posed a challenge handing time and numeric resources. This necessitated the devel-
opment of a modelling language capable of expressing temporal and numeric properties
of planning domains.
In an original paper, Long and Fox (2003) describe the language, PDDL2.1, which
was used in the competition. They describe the syntax of the language, its formal
semantics and the validation of concurrent plans. They reported that PDDL2.1 has
considerable modelling power - exceeding the capabilities of current planning technology
- and presents a number of important challenges to the research community. Clearly, this
paper generated a series of commentaries that suggest that Long and Fox’s contribution
(2003) would stimulate active research that may keep researchers busy for decades. The
following are some of the reactions to the novel contribution by Long and Fox (2003).
In a commentary paper reacting to the proposal by Long and Fox (2003), Bacchus
(2003) argue that although PDDL is a very useful standard for the planning competi-
tion, but its design does not properly consider the issue of domain modelling. Hence,
the critic did not advocate its use in specifying planning domains outside of the context
of the planning competition. The author states that the ﬁeld needs to explore diﬀer-
ent approaches and grapple more directly with the problem of eﬀectively modeling and
utilising all of the diverse pieces of knowledge we typically have about planning domains.
In another reaction on PDDL, Boddy (2003) comments that it was originally con-
ceived and constructed as a lingua franca for the International Planning Competition,
and that PDDL2.1 embodies a set of extensions intended to support the expression of
something closer to “real planning problems.” The author states that this objective has
only been partially achieved due in large part to a deliberate focus on not moving too
far from classical planning models and solution methods.
Geﬀner (2003) comments on the PDDL 2.1 language and its use in the planning
competition, focusing on the choices made for accommodating time and concurrency.
The author also discusses some methodological issues that have to do with the move
toward more expressive planning languages and the balance needed in planning research
between semantics and computation.
“M19N41” — 2008/11/5 — 17:14 — page 552 — #18 i
552 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
McDermott (2003) notes that PDDL2.1 was designed to push the envelope of what
planning algorithms can do, and it has succeeded. It adds two important features: dura-
tive actions, which take time (and may have continuous eﬀects); and objective functions
for measuring the quality of plans. The concept of durative actions is ﬂawed; and the
treatment of their semantics reveals too strong an attachment to the way many con-
temporary planners work. Future PDDL innovators should focus on producing a clean
semantics for additions to the language, and let planner implementers worry about cou-
pling their algorithms to problems expressed in the latest version of the language.
Smith (2003) comments that the addition of durative actions to PDDL2.1 sparked
some controversy. Fox and Long argued that actions should be considered as instanta-
neous, but can start and stop processes. Ultimately, a limited notion of durative actions
was incorporated into the language. Smith (2003) argues that this notion is still im-
poverished, and that the underlying philosophical position of regarding durative actions
as being a shorthand for a start action, process, and stop action ignores the realities of
modelling and execution for complex systems.
With the standard language PDDL2.1, Gerevini et al. (2003) present some tech-
niques for planning in domains speciﬁed that supports ‘durative actions’ and numerical
quantities. These techniques are implemented in LPG, a domain-independent planner
that took part in the 3 rd International Planning Competition (IPC). LPG is an incre-
mental, any time system producing multi-criteria quality plans. The core of the system
is based on a stochastic local search method and on a graph-based representation called
‘Temporal Action Graphs’ (TA-graphs).
The paper by Gerevini et al. (2003) focuses on temporal planning, introducing TA-
graphs and proposing some techniques to guide the search in LPG using this represen-
tation. The experimental results of the 3 rd IPC, as well as further results presented in
the paper, show that the techniques can be very eﬀective. Often LPG outperforms all
other fully automated planners of the 3 rd IPC in terms of speed to derive a solution, or
quality of the solutions that can be produced. SAPA is another planner that has been
developed for the manufacturing environment.
SAPA is a domain-independent heuristic forward chaining planner that can handle
durative actions, metric resource constraints, and deadline goals. It is designed to be ca-
pable of handling the multi-objective nature of metric temporal planning. The technical
contributions of Do and Kambhampati (2003) in the development of SAPA include (i)
planning-graph based methods for deriving heuristics that are sensitive to both cost and
“M19N41” — 2008/11/5 — 17:14 — page 553 — #19 i
A Literature Review on Artiﬁcial Intelligence 553
makespan (ii) techniques for adjusting the heuristic estimates to take action interactions
and metric resource limitations into account and (iii) a linear time greedy post-processing
technique to improve execution ﬂexibility of the solution plans. An implementation of
SAPA using many of the techniques presented in their paper was one of the best domain
independent planners for domains with metric and temporal constraints in the third In-
ternational Planning Competition, held at AIPS-02. Their paper describes the technical
details of extracting the heuristics and presents an empirical evaluation of the current
implementation of SAPA.
The work by Nau et al. (2003) explored the features of SHOP 2 from an HTN
planning systems perspective. In particular, the features of SHOP 2 which enabled
to excel in the international conference on AI planning and schedule (AIPs) in 2002
competitions, especially those aspects of SHOP 2 that deals with temporal and metric
planning domains were explored.
In Nebel (2000) the compilability and expressive power of prepositional planning
formalism were studied. The author formalise the intuition that the expressive power is
a measure of how concisely planning domains and plans can be expressed in a partic-
ular formalisms by introducing the notion of “compilations schemes” between planning
formalisms. The authors analysed the expressiveness of a large family of prepositional
planning formalisms ranging from basic STRIPS to formalism with conditional eﬀects,
partial state speciﬁcations, and prepositional formulae in the preconditions. The result
conﬁrms that the proposed extensions to the CRAPU PLAN algorithm concerning condi-
tional eﬀects are optimal with respect to the “compilability” framework. Another result
is that general proportional formulae cannot be compiled into conditional eﬀects of the
Jessen and Veloso (2000) introduce a new planning domain description language,
NADL, to specify non-deterministic, deﬁnition of controllable agents and uncontrollable
environment agents. They present empirical result applying UMOP to domain ranging
from deterministic and single agent with no environment actions. UMOP is shown to be
a rich and eﬃcient planning system.
In the paper by Brafman (2001) the author compare the ability of two classes of
algorithms to propagate and discover reachability and relevance constraints in classi-
cal planning problems. The results shed light on the ability of diﬀerent plan-encoding
schemes to propagate information forward and backward, and in the relative merit of
plan-level and SAT-level pruning methods.
“M19N41” — 2008/11/5 — 17:14 — page 554 — #20 i
554 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
Boutilier and Brafman (2001) investigated into partial-order planning with concur-
rent interacting actors. The authors demonstrate this fact by developing a sound and
complete partial-order planner for planning with concurrent. Interacting actions, POMP,
that extends the existing partial-order planners in a straightforward way. These results
open way to the use of partial-order planners for the centralised control of cooperative
In another article, Ambite and Knoblock (2001) introduces planning by Rewriting
(PBR), a new paradigm for eﬃcient high-quality domain-independent planning. PBR
exploits declarative plan-rewriting rules and eﬃcient local search-techniques to transform
an easy-to-regenerate, but possibly suboptimal, initial plan into a high-quality plan. The
experimental results show that the PBR approach provides signiﬁcant savings in planning
eﬀort while generating high-quality plans.
3.2. Applications of AI in Robots
Robots are advanced automation technologies generally used for production and
non-production activities in order to make life easier and to improve productivity at the
work place. In the manufacturing systems, many manufacturers have turned robotics and
automation for more reliable manufacturing system solutions. Application examples of
robots are found in the construction industry, car parks, nuclear installations, airports,
mines, hospitals, welding shipyards, space stations, and automotive applications. In
particular, robots are found in unusual places where the environmental and working
conditions presents hazards and/or where dangerous tasks are performed.
Studies on robots have been viewed from three dimensions - its navigation, robot’s
localisation and robot’s participation in agent teams. The study that deals with robot
navigation is credited to Shatkay and Kaelbling (2002). The authors describe a formal
framework for incorporating readily available odometric information and geometrical
constraints into both the model and the algorithm that learns them.
In another work, Fox et al. (1999) present a version of Markov localisation which pro-
vides accurate position estimates and which is tailored towards dynamic environments.
The key idea to Markov localisation is to maintain a probability density over the space of
all locations of a robot in its environment. The work here presents an entirely diﬀerent
environment from what obtains in the literature. Thus, it is a unique contribution to
knowledge. Robots was linked to agent teams in an execution monitoring approach used
to implement execution assistants (EAs) in two diﬀerent dynamic, data-rich, real-world
“M19N41” — 2008/11/5 — 17:14 — page 555 — #21 i
A Literature Review on Artiﬁcial Intelligence 555
domains to assist human in monitoring team behaviour. The credit of the approach lies
in that it customises monitoring behaviour for each speciﬁc task, plan and situation, as
well as for user preferences.
3.3. Applications of AI in general
The study by Franntz (2003) is an example in this respect. The focus is the use of
AI as a framework for understanding intuition. His research shows how the overlaps in
Herbert Simon’s work and especially his work on AI aﬀected his view towards intuition.
Hebert Simon made overlapping substantive contributions to the ﬁelds of economics,
psychology, cognitive science, AI, decision theory, and organisation theory.
Simon’s work was motivated by the belief that the human mind, human thinking,
decision-making in man, and human creativity need to be mysterious. It was after
he helped create “thinking” machines that Simon came to understand human intuition
as subconscious pattern recognition. In doing so he showed that intuition need not
be associated with magic and mysticism, and that it is complementary with analytical
A related work to the work of Franntz (2003) is the work presented by Alai (2004).
His work centers on whether scientiﬁc discovery is a rational and logical process. If it is,
according to the AI hypothesis, it should be possible to write computer programs able
to discover laws or theories; and if such programs were written, this would deﬁnitely
prove the existence of logic of discovery. However, a program written by a Simon led
group according to this line of reasoning proved abortive. The program was able to infer
famous laws of physics and chemistry; but having found no new law, it could not be
exactly called a discovery program. The programs written in the ’Turning tradition’,
instead, produced new and useful empirical generalisation, but no theoretical discovery.
Thus failing to move the logical character of the most signiﬁcant kind of discoveries.
A new cognitivist and connectionist approach by Holland, Holy oak, Nisbett and
Thagard, hooks more promisingly. A study of their proposals helps to understand the
complex character of discovery processes, the abandonment of belief in the logic of dis-
covery by logical positivists and the necessity of a realist interpretation of scientiﬁc
Similarly, Franklin (2003) deals with the representation of context with ideas drawn
from AI. To move beyond vague platitudes about the importance of a context in le-
gal reasoning or natural language understanding, one must take account of ideas from
“M19N41” — 2008/11/5 — 17:14 — page 556 — #22 i
556 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
AI on how to represent context formally. Work on topics like prior probabilities, the
theory-laderness of observation, encyclopedic knowledge for disambiguation in language
translation and pathology text diagnosis has produced a body of knowledge on how to
represent context in AI applications.
In another work, Haynes (2003) reviews the shift from Alchemy to AI in western
literature. The simpliﬁcation underlying contemporary mythology of knowledge arises
from fear of the power change that science entails leaving many people feeling confused
and disempowered. Kit reemerges in the media, most often under the name of “Franken-
stein”, without any new discovery that appears to threaten social equilibrium. This is
not a new phenomenon. From medieval stories about alchemists to ﬁlms about computer
hackers, good scientists are in the minority, and the number of recurring stereotypes is
small. These archetypes oﬀer writers and ﬁlmmakers’ convenient shorthand, a matrix
in which to slot contemporary scientists and their projects, simplifying the issues. Like
all myths, they appear simple but represent complex ideas and suppressed fears, which
transcend time, place, and race.
Further work on general studies on AI could be credited to Holte and Choueiry
(2003) and Zucker (2003). These authors worked on the abstraction concept in AI. In
the case of Holte and Choueiry (2003), the authors contributed in two ways to the aims
of the special issue on abstraction. The ﬁrst is to show that there are compelling reasons
motivating the use of abstraction in the purely computational realm of AI.
The second is to contribute to the overall discussion of the nature of abstraction
by providing examples of the abstraction process currently used in AI. Although each
type of abstraction is speciﬁc to a somewhat narrow context, it is hoped that collectively
they illustrate the richness and variety of abstraction in its fullest sense. Furthermore,
Zucker (2003) used abstraction to account for the use of various levels of details in a given
representation language or the ability to change from one level to another while preserving
useful properties. Abstraction has been mainly studied in problem solving, theorem
proving, knowledge representation (in particular for spatial and temporal reasoning) and
In such contexts, abstraction is deﬁned as a mapping between formalisms that re-
duces the computational complexity of the task at stake. By analysing the notion of
abstraction from an information quantity point of view, we pinpoint the diﬀerences and
the complementary role of reformulation and abstraction in any representation change.
The author contributes to extending the existing semantic theories of abstraction to be
“M19N41” — 2008/11/5 — 17:14 — page 557 — #23 i
A Literature Review on Artiﬁcial Intelligence 557
grounded or perception, where the notion of information quantity is easier to charac-
terise formally. In the author’s view, abstraction is best-represented using abstraction
operators, as they provide semantics for classifying diﬀerent abstractions and support the
automation of representation changes. The usefulness of a grounded theory of abstraction
in the cartography domain is illustrated. Finally, the importance of explicitly represent-
ing abstraction for designing more autonomous and adaptive system is discussed.
In another study, Kim (1995) considers whether Godel’s results preclude the possi-
bility or the impossibility of the AI thesis; and also what the (possible) applications or
consequences of them are for AI research. The author shows that while the limitative
Godel’s results are shown to preclude neither the possibility nor the impossibility of AI
thesis, they have and will continue to shed signiﬁcant light on the development of the AI
Also under general studies of AI, the account given by Sivramkrishna and Panigrahi
(2003) present the concept as a tool in development planning. In particular the Kohonen
self-organising map, is a user-friendly tool for development planners and practitioners to
explore patterns in development. An application with several indicators over 399 Indian
districts illustrates the need to study development patterns. The paper also makes clear
the versatility of the kohonen self-organising map technique in exploring these regional
patterns of development. In another paper, the concept of application of AI in short
term electric load forecasting was considered (Metaxiotis et al., 2003).
The paper provides an overview for the researcher of AI technologies, as well as their
current use in the ﬁeld of short-term electric load forecasting (STELF). The history of
AI in STELF is outlined, leading to a discussion of the various approaches as well as the
current research directions. The paper concludes by sharing thoughts and estimations
on AI future prospects in this area. The review reveals that although still regarded as a
novel methodology; AI technologies are shown to have matured to the point of oﬀering
real practical beneﬁts in many of their applications. Still under general considerations
the future of AI is considered by Clocksin (2003). The author considers some of the ideas
inﬂuencing current AI research and outlines an alternative conceptual framework that
gives priority to social relationships as a key component and constructor of intelligent
The framework starts from Weizenbaum’s observation that intelligence manifests
itself only relative to speciﬁc social and cultural contexts. This is in contrast to a
prevailing view, which sees intelligence as an abstract capability of the individual mind
“M19N41” — 2008/11/5 — 17:14 — page 558 — #24 i
558 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
based on a mechanism for rational thought. The new approach is not based on the
conventional idea that the mind is a rational processor of symbolic information, nor does
it require the idea that thought is a kind of abstract problem solving with a semantics
that is independent of its embodiment. Instead, priority is given to aﬀective and social
responses that serve to engage the whole agent in the life of the communities in which it
Intelligence is seen not as the deployment of capabilities for problem solving, but as
constructed by the continual, ever-changing and unﬁnished engagement with the social
group within the environment. The constructions of the identity of the intelligent agent
involve the appropriation of ’taking up’ of positions within the conversations and nar-
ratives in which it participates. Thus, the new approach argues that intelligent agent
is shaped by the meaning ascribed to experience, by its situation in the social matrix
and by practices of self and of relationship into which intelligent life is recruited. This
has implications for the technology of the future, for example, classic AI models such as
goal-directed problem solving are seen as special cases of narrative practices instead of
as ontological foundations.
Yet in another work, the paper by Schiaﬀonati (2003) aims to put the basis of
an extended and well-founded philosophy of AI: it delineates a multi-layered general
framework to which diﬀerent contributions in the ﬁeld may be traced back. The core
point is to underline how in the same scenario both the role of philosophy on AI and
role of AI on philosophy must be considered. Moreover, this framework is revised and
extended in the light of the consideration of a type of multiagent system devoted to
aﬀord the issue of scientiﬁc discovery both from a conceptual and from a practical point
The paper by Moraga et al. (2003) reviews one particular area of AI, which roots may
be traced back to Multiple-valued Logic: the area of fuzzy control. After an introduction
based on an experimental scenario, basic cases of fuzzy control are presented and formally
analysed. Their capabilities are discussed and their constraints are explained. Finally
it is shown that a parameterisation of either the fuzzy sets or the connectives used to
express the rules governing a fuzzy controller allows the use of new optimisation methods
to improve the overall performance.
In concluding this section, we point out that there is a vast array of studies that
have not been considered yet but fall under the above groupings. Such studies can be
“M19N41” — 2008/11/5 — 17:14 — page 559 — #25 i
A Literature Review on Artiﬁcial Intelligence 559
found in (Zhang and Zhang (2001), Neal (2000), Kaminka and Tambe (2000), Siskind
(2001), Walker et al. (2002), Tan et al. (2003) and Ygge and Akkermans (1999)).
3.4. Artiﬁcial intelligence in the manufacturing ﬁeld
In the manufacturing ﬁeld prominent research has been carried out in a number of
areas, including quality monitoring and production scheduling, among others.
In a study, Stefanuk and Zhozhikashvili (2002) carried out an analysis of the pro-
duction and rules in the way they are used in AI systems. The proposed new deﬁnition
for productions refers to a large number of types of production that may be found in
the literature on AI systems. This deﬁnition emphasises in the most general way those
production components that are important both for theory and for practice and which
for some reasons remained unnoticed by many researchers. These components are sup-
plemented in a theoretical formalism that concludes the paper.
In the area of manufacturing, Toni et al. (1996) proposed an artiﬁcial, intelligence
- based production scheduler. The production scheduler ultilises a hybrid push/pull
approach to scheduling and exploits the expert system technology in order to obtain
satisfactory solutions. The scheduler is applied to a multi-stage production and inventory
system, managed by make-to-order, with a large variety of incoming orders. The search
for solution is made in respect of the due-dates and under eﬃciency constraints (minimum
lot maximum storehouse levels e.t.c.). The work considers order aggregation, both a
portfolio and production level. Provides a dynamic rescheduling mechanism. It outlines
theoretical arguments in favour of the scheduler and notes practical advantages as a
consequence of the application of the scheduler in a ﬁrm, which utilised a traditional
Another interesting research was carried out by Bhuyanb (2003) on tea quality pre-
diction using a tin oxide-based electronic nose with an AI approach. In the research, the
authors analysed using a metal oxide sensor (MOS) based electronic nose (EN) ﬁve tea
samples with diﬀerent qualities: normally drier month, drier mouth again over ﬁred, well
fermented normal dried in oven, well-fermented over-ﬁred in oven and under-fermented
normal ﬁred in oven. Mainly its taste and smell determine the ﬂavour of tea, which are
determined by hundreds of volatile organic compounds (VOC) and non-volatile organic
compounds present in tea. Tea ﬂavour is traditionally measured through the use of a
combination of conventional analytical instrumentation and human organoleptic proﬁl-
ing panels. These methods are expensive in terms of, for example, time and labour.
“M19N41” — 2008/11/5 — 17:14 — page 560 — #26 i
560 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
The methods are also inaccurate because of a lack of either sensitivity or quantitative
An investigation was made to determine the ﬂavour of diﬀerent tea samples using
EN and thus to explore the possibility of replacing existing analytical and proﬁling panel
methods. The technique uses an array of four mosses, each of, which has an -electrical
resistance that has partial sensitivity to the headspace of tea. The signals from the
sensor array are then conditioned by suitable interface circuitry resulting in the tea
data set. The data were processed using principal component analysis (PCA), fuzzy C
means (FCM) algorithm. The data were then analysed following the neural network
paradigms, following the self-organising map (SOM) method along with radial basis
function (RBF) network and probabilistic neural network (PNN) classiﬁer, using FCM
and SOM feature extraction techniques along with RBF neural network. We achieved
100% correct classiﬁcation for the ﬁve diﬀerent tea samples, each of which has diﬀerent
qualities. These results prove that the EN is capable of discriminating between the
ﬂavours of teas manufactured under diﬀerent processing condition, viz. over-fermented,
over-ﬁred, under-fermented, etc.
3.5. Artiﬁcial intelligence in maintenance
Another group of studies in AI is in the area of maintenance where a set of research
that focuses on the maintenance of systems whose output is tangible. The other set is
concerned primarily with design of maintenance systems for intangible products. A good
example of this could be drawn from wrapper maintenance. Wrappers are intangible
outputs in web sources with the function of extracting data.
Diez et al. (2002) investigated into an AI approach for improving plant operator
maintenance proﬁciency. The aspect of linkage of construction plant maintenance prac-
tice and its plant operators are the central focus. Draw from the knowledge that unlike
plant operating within the manufacturing sector, construction plant is seen as largely
dependent upon operator skill and competence to maintain the item in a safe, fully
Research has previously successfully modeled machine breakdown, but revealed that
the operator’s impact upon machine breakdown rates can be considerable. A conceptual
model methodology with which to assess the maintenance proﬁciency of individual plant
operators was therefore presented by Diez (2002). In the aspect of condition monitoring
“M19N41” — 2008/11/5 — 17:14 — page 561 — #27 i
A Literature Review on Artiﬁcial Intelligence 561
AI has been used as a device for monitoring the condition of tools in an engineering
3.6. Applications of AI in environmental pollution
It is interesting to note that AI has been widely used in many aspects of human
lives. A good case is presented by Chan et al. (2003). The authors stated that AI could
be applied to the reduction of environmental pollution, conservation and recycling since
natural resources are signiﬁcant social and environmental concerns. As valuable means
for pollution control, they noted that minimisation and mitigation remain attractive
approaches. However, interactive, dynamic and uncertain features are associated with
these processes, resulting in diﬃculties in their management and control. AI is consid-
ered as an eﬀective approach for tackling these complexities. Their study examines the
recent advancements of AI-based technologies for management and control of pollution
minimisation and mitigation processes.
In the area of environmental pollution, AI has been used for management and control
of pollution minimisation and mitigation processes. The literature relevant to the area of
application of AI to control and management of pollution minimisation and mitigation
processes were investigated, especially, technologies of expert systems, Fuzzy logic, and
neural networks, which emerge as the most frequency employed approaches for realising
process control, and are highlighted. The results not only provide an overview of the
updated progress in the study ﬁeld but also, more importantly, reveal perspectives of
research for more eﬀective environmental process control through the AI-aided measures.
Several demanding areas for enhanced research eﬀorts are discussed, including issues of
data availability and reliability, methodology validity, and system complexity.
4. General Remarks and Future Directions
This paper began with a realisation that we are in a wonderful age of discovery about
issues concerning AI. A number of impressive documentations of established research
methods and philosophy have been discussed in print for several years. Unfortunately,
little comparison and integration across studies exist. In this article, we have set out
to create a common understanding of AI research. As much as it is the goal to declare
about the purpose of writing this paper, it is important for us to declare about what the
purpose is not.
“M19N41” — 2008/11/5 — 17:14 — page 562 — #28 i
562 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
This paper is not attempting to provide an all-encompassing framework on the liter-
ature on AI research. Rather, it is attempting to provide a starting point for integrating
knowledge across research in this domain and suggest avenues of future research. It
explored studies in certain novel disciplines: environmental pollution, medicine, mainte-
nance, manufacturing, etc.
Further research is needed to extend the present frontier of knowledge in AI by
integrating principles and philosophies of some traditional disciplines into the existing AI
frameworks Markham et al., 2000. For example, in designing an AI system for a medical
issue on a survey-based study, the principles of statistical signiﬁcance, conﬁdence limit,
experimental design and hypothesis testing may improve the value of the research and
the output of the software designed.
In the area of agriculture, the use of AI may be made in the validity of bovine car-
cass classiﬁcation. AI algorithms could be tested to assess possible diﬀerences in the
behaviour of the classiﬁers in aﬀecting the repeatability of grading. With this devel-
opment, clariﬁcation and standardisation of the beef market in various countries and
regions could be made. In addition, since conformation of light and standard carcasses
can be considered to be of diﬀerent traits, this could improve carcass conformation scores
from markets, thus presenting a great variety of ages and weights of slaughtered animals.
AI search techniques is important for the circuits and systems design space explo-
ration. Future researchers could explain what sorts of search techniques are useful for
this aim. Again, the place, role and way of use of these techniques in circuit and system
design could be investigated. Search techniques such as heuristics for the automatic con-
struction and selection of the most promising solutions to the circuit synthesis problems
could be developed.
AI methods could be applied to estimate tool wear in lathe turning. Use could be
made of conventional AI methods, neural network, and the fuzzy decision support system.
An important variable to study is the tool wear estimation is based on the measurement
of cutting force components.
While we do not expect this paper to spark a sudden proliferation of an already
established ﬁeld, we believe that this research can be an important intellectual tool for
both the refocusing of the work and the creating new intellectual opportunities. This
paper presents valuable ideas and perspectives for undergoing research on AI. As stated
earlier, research related to AI has proliferated in recent years. We do not pretend to be
“M19N41” — 2008/11/5 — 17:14 — page 563 — #29 i
A Literature Review on Artiﬁcial Intelligence 563
inclusive to deceive ourselves that any of these ideas represent the thinking of all, or even
From a search of relevant literature, these were the themes that emerged and sur-
faced, not only in computer journals but also across a range of scientiﬁc journals. We
anticipate the transformation of the discipline in future age. This will be a journey that
may experience change in its course as new generations of scholars contribute to the
dialogue and to the action. As noted earlier, this work presents a review, hence, it lays a
foundation for future inquiry. It has not only oﬀered a basis for future comparisons, but
has prompted a number of new questions for investigations as well. While topics that
might be considered as results of this work are numerous, some are of particularly broad
interest or impact.
 Acid, S. and De Campos. L. M., Searching for bayesian network structures in the space of restricted
acyclic partially directed graphs, Journal of Artiﬁcial Intelligence Research, Vol. 18, pp.445-490,
 Alai, M., AI, scientiﬁc discovery and realism, Minds and Machines, Vol. 14, No. 1, pp.21-42, 2004.
 Al-Ani, A. and Deriche, M., A new technique for combining multiple classiﬁers using the dempster-
shafer theory of evidence, Journal of Artiﬁcial Intelligence Research, Vol. 17, pp.333-361, 2002.
 Ambite, J. L. and Knoblock, C. A., Planning by rewriting, Journal of Artiﬁcial Intelligence Research,
Vol. 15, pp.207-261, 2001.
 Andrew, A. M., Short paper: lifemanship and AI. Kybernetes: The International Journal of Systems
& Cybernetics, Vol. 30, No. 2., 2001.
 Argamon-Engelson, S. and Dagan, I., Committee-based sample selection for probabilistic classiﬁers,
Journal of Artiﬁcial Intelligence Research, Vol. 11, pp.335-360, 1999.
 Atkinson K. and Bench-Capon T., Practical reasoning as presumptive argumentation using action
based alternating transition systems, Artiﬁcial Intelligence, Vol. 171, No. 10-15, pp.855-874, 2007.
 Baader, F., Lutz, C., Sturm, H. and Wolter, F., Fusions of description logics and abstract description
systems, Journal of Artiﬁcial Intelligence Research, Vol. 16, pp.1-58, 2002.
 Bacchus, F., The power of modeling - a response to PDDL2.1., Journal of Artiﬁcial Intelligence
Research, Vol. 20, pp.125-132, 2003.
 Baget, J. F. and Mugnier, M. L., Extensions of simple conceptual graphs: the complexity of rules
and constraints, Journal of Artiﬁcial Intelligence Research, Vol. 16, pp.425-465, 2002.
 Balazinski, M., Czogala, E., Jemielniak, K. and Leslie J., Tool condition monitoring using artiﬁcial
intelligence methods, Engineering Applications of Artiﬁcial Intelligence, Vol. 15, Issue 1, pp.73-80,
 Barber, F., Reasoning on Interval and Point-based Disjunctive Metric Constraints in temporal con-
texts, Journal of Artiﬁcial Intelligence Research, Vol. 12, pp.35-86, 2000.
 Barzilay, R., Elhadad, N. and McKeown K. R., Inferring strategies for sentence ordering in mul-
tidocument news summarization, Journal of Artiﬁcial Intelligence Research, Vol. 17, pp.s 35-55,
“M19N41” — 2008/11/5 — 17:14 — page 564 — #30 i
564 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
 Basu, C., Hirsh, H., Cohen, W. W. and Nevill-Manning, C., Technical paper recommendation: a
study in combining multiple information sources, Journal of Artiﬁcial Intelligence Research, Vol. 14,
pp.231-252, 2001. s
 Baxter, J., A model of inductive bias learning, Journal of Artiﬁcial Intelligence Research, Vol. 12,
 Baxter, J. and Bartlett, P. L., Inﬁnite-horizon policy-gradient estimation, Journal of Artiﬁcial Intel-
ligence Research, Vol. 15, pp.319-350, 2001.
 Baxter, J., Bartlett, P. L. and Weaver, L., Experiments with inﬁnite-horizon, policy-gradient Esti-
mation, Journal of Artiﬁcial Intelligence Research, Vol. 15, pp.351-381, 2001.
 Becker, A., Bar-Yehuda, R. and Geiger, D., Randomised algorithms for the loop cutset problem,
Journal of Artiﬁcial Intelligence Research, Vol. 12, pp.219-234, 2000.
 Bhattacharyya, C. and Keerthi, S. S., Mean ﬁeld methods for a special class of belief networks,
Journal of Artiﬁcial Intelligence Research, Vol. 15, pp.91-114, 2001.
 Bhuyanb, Tea quality prediction using a tin oxide-based electronic nose: an artiﬁcial intelligence
approach, Sensors and Actuator B: Chemical, Vol. 94, Issue 2, pp.228-237, 2003.
 Birnbaum, E. and Lozinskii, E. L., The good old Davis-Putnam procedure helps counting models,
Journal of Artiﬁcial Intelligence Research, Vol. 10, pp. 457-477, 1999.
 Blockeel, H., Dehaspe, L., Demoen, B., Janssens, G., Ramon, J. and Vandecasteele, H., Improving
the eﬃciency of inductive logic programming through the use of query packs, Journal of Artiﬁcial
Intelligence Research, Vol. 16, pp.135-166, 2002.
 Bod, R., A uniﬁed model of structural organisation in language and music, Journal of Artiﬁcial
Intelligence Research, Vol. 17, pp.289-308, 2002.
 Boddy, M. S., Imperfect match: PDDL 2.1 and real applications, Journal of Artiﬁcial Intelligence
Research, Vol. 20, pp.133-137, 2003.
 Boutilier, C. and Brafman, R. I., Partial-order planning with concurrent interacting actions, Journal
of Artiﬁcial Intelligence Research, Vol. 14, pp.105-136, 2001.
 Boutilier, C., Dean, T., Hanks, S., Decision-theoretic planning: Structural assumptions and compu-
tational leverage, Journal of Artiﬁcial Intelligence Research, Vol. 11, pp. 1-94, 1999.
 Brafman, R. I., On reachability, relevance, and resolution in the planning as satisﬁability approach,
Journal of Artiﬁcial Intelligence Research, Vol. 14, pp.1-28, 2001.
 Brafman, R. I. and Domshlak, C., Structure and complexity in planning with unary operators, Journal
of Artiﬁcial Intelligence Research, Vol. 18, pp.315-349, 2003.
 Brafman, R. I. and Tennenholtz, M., Learning to coordinate eﬃciently: A model-based approach,
Journal of Artiﬁcial Intelligence Research, Vol. 19, pp.11-23, 2003.
 Brodley, C. E. and Friedl, M. A., Identifying mislabeled training data, Journal of Artiﬁcial Intelligence
Research, Vol. 11, pp.131-167, 1999.
 Brooks, R. A., Cambrian Intelligence: The early history of the new AI. Kyberntes: The International
Journal of Systems & Cybernetics, Vol. 39 No. 1., 2001.
 Bui, H. H., Venkatesh, S. and West, G., Policy recognition in the abstract hidden markov model,
Journal of Artiﬁcial Intelligence Research, Vol. 17, pp.451-499, 2002.
 Cadoli, M., Donini, F. M., Liberatore, P. and Schaerf, M., Space eﬃciency of propositional knowledge
representation formalisms, Journal of Artiﬁcial Intelligence Research, Vol. 13, pp.1-31, 2000.
 Calvanese, D., Lenzerini, M. and Nardi, D., Unifying class-based representation formalisms, Journal
of Artiﬁcial Intelligence Research, Vol. 11, pp.199-240, 1999.
 Cemgil, A. T. and Kappen, B., Monte carlo methods for tempo tracking and rhythm quantization,
Journal of Artiﬁcial Intelligence Research, Vol. 18, pp.45-81, 2003.
 Chan, H. and Darwiche, A., When do numbers really matter?, Journal of Artiﬁcial Intelligence
Research, Vol. 17, pp.265-287, 2002.
“M19N41” — 2008/11/5 — 17:14 — page 565 — #31 i
A Literature Review on Artiﬁcial Intelligence 565
 Chan, C. W. and Huangb, G. H., Artiﬁcial intelligence for management and control of pollution
minimisation and mitigation processes. Engineering Applications of Artiﬁcial Intelligence, Vol. 16
Issue 2, pp.75-90, 2003.
 Chawla, N. V., Bowyer, K. W., Hall, L. O. and Kegelmeyer, W. P., SMOTE: Synthetic minority
over-sampling technique, Journal of Artiﬁcial Intelligence Research, Vol. 16, pp.321-357, 2002.
 Chen, X. and Van Beek, P., Conﬂict-directed backjumping revisited, Journal of Artiﬁcial Intelligence
Research, Vol. 14, pp.53-81, 2001.
 Cheng, J. and Druzdzel, M. J., AIS-BN: An adaptive importance sampling algorithm for evidential
reasoning in large Bayesian networks, Journal of Artiﬁcial Intelligence Research, Vol. 13, pp.155-188,
 Cimatti, A. and Roveri, M., Conformant planning via symbolic model checking, Journal of Artiﬁcial
Intelligence Research, Vol. 13, pp.305-338, 2000.
 Clocksin, W. F., Artiﬁcial intelligence and the future. Philosophical Transactions: Mathematical,
Physical & Engineering Sciences, Vol. 361, No. 1809, pp.1721-1748, 2003.
 Console, L., Picardi, C. and Theseider Dupre, D., Temporal decision trees: Model-based diagnosis of
dynamic systems on-board, Journal of Artiﬁcial Intelligence Research, Vol. 19, pp.469-512, 2003.
 Cristani, M., The complexity of reasoning about spatial congruence, Journal of Artiﬁcial Intelligence
Research, Vol. 11, pp.361-390, 1999.
 Darwiche, A. and Marquis, P., A knowledge compilation map, Journal of Artiﬁcial Intelligence Re-
search, Vol. 17, pp.229-264, 2002.
 Debruyne, R. and Bessiere, C., Domain ﬁltering consistencies, Journal of Artiﬁcial Intelligence
Research, Vol. 14, pp.205-230, 2001.
 Dietterich, T. G., Hierarchical reinforcement learning with the MAXQ value function decomposition,
Journal of Artiﬁcial Intelligence Research, Vol. 13, pp.227-303, 2000.
 Diez, J., Bahamonde, A., Alonso, J., Lopez, S., Del Coz, J. J., Quevedo, J. R., Ranilla, J., Luaces,
O., Alvarez, I., Royo, L. J., Dutta, R., Hinesa, E. L., Gardnera, J. W., Kashwanb, K. R., Edwards,
D. J., Holt, G. D. and Robinson, B., An artiﬁcial intelligence approach for improving plant operator
maintenance proﬁciency. Journal of Quality in Maintenance Engineering; Vol. 8 No. 3., 2002.
 Di Sciascio, E., Donini, F. M. and Mongiello, M., Structured know ledge representation for image
retrieval, Journal of Artiﬁcial Intelligence Research, Vol. 16, pp.209-257, 2002.
 Do, M. and Kambhampati, S., SAPA: A multi-objective metric temporal planner, Journal of Artiﬁcial
Intelligence Research, Vol. 20, pp.155-194, 2003.
 Drummond, C., Accelerating reinforcement learning by composing solutions of automatical ly identi-
ﬁed subtasks, Journal of Artiﬁcial Intelligence Research, Vol. 16, pp.59-104, 2002.
 Edelkamp, S., Taming numbers and durations in the model checking integrated planning system,
Journal of Artiﬁcial Intelligence Research, Vol. 20, pp.195-238, 2003.
 Eiter, T., Faber, W., Leone, N., Pfeifer, G. and Polleres, A., Answer set planning under action costs,
Journal of Artiﬁcial Intelligence Research, Vol. 19, pp.25-71, 2003.
 Elomaa, T. and Kaariainen, M., An analysis of reduced error pruning, Journal of Artiﬁcial Intelli-
gence Research, Vol. 15, pp.163-187, 2001.
 Fern, A., Givan, R. and Siskind, J. M., Speciﬁc-to-General learning for temporal events with appli-
cation to learning event deﬁnitions from video, Journal of Artiﬁcial Intelligence Research, Vol. 17,
 Finkelstein, L., Markovitch, S., Rivlin, E., Optimal schedules for parallelising anytime algorithms:
The case of shared resources, Journal of Artiﬁcial Intelligence Research, Vol. 19, pp. 73-138, 2003.
 Fox, D., Burgard, W. and Thrun, S., Markov localisation for mobile robots in dynamic environments,
Journal of Artiﬁcial Intelligence Research, Vol. 11, pp.391-427, 1999.
“M19N41” — 2008/11/5 — 17:14 — page 566 — #32 i
566 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
 Fox, M. and Long, D., PDDL2.1: An Extension to PDDL for Expressing Temporal Planning Do-
mains, Journal of Artiﬁcial Intelligence Research, Vol. 20, pp.61-124, 2003.
 Franklin, J., The representation of context: ideas from artiﬁcial intelligence. Law, Probability and
Risk, Vol. 2, No. 3, pp.191-199, 2003.
 Franntz, R., Herbert Simon - Artiﬁcial intelligence as a framework for understanding intuition. The
Economic Psychology of Herbert A. Simon, pp.265-277, 2003.
 Gamberger, D. and Lavrac, N., Expert-guided subgroup discovery: Methodology and application,
Journal of Artiﬁcial Intelligence Research, Vol. 17, pp.501-527, 2002.
 Gao, Y. and Culberson, J., An analysis of phase transition in NK landscapes, Journal of Artiﬁcial
Intelligence Research, Vol. 17, pp.309-332, 2002.
 Geﬀner, H. A., PDDL 2.1: Representation vs. computation, Journal of Artiﬁcial Intelligence Re-
search, Vol. 20, pp.139-144, 2003.
 Gerevini, A., Saetti, A. and Serina, I., Planning through stochastic local search and temporal action
graphs in LPG, Journal of Artiﬁcial Intelligence Research, Vol. 20, pp.239-290, 2003.
 Ginsberg, M. L., GIB: Imperfect information in a computational ly challenging game, Journal of
Artiﬁcial Intelligence Research, Vol. 14, pp.303-358, 2001.
 Gordon, D. F., Asimovian adaptive agents, Journal of Artiﬁcial Intelligence Research, Vol. 13,
 Goyache, F., Artiﬁcial intelligence techniques point out diﬀerences in classiﬁcation performance be-
tween light and standard bovine carcasses. Meat Science, Vol. 64, Issue 3, pp.219-331, 2003.
 Grumberg, O., Livne, S. and Markovitch, S., Learning to order BDD variables in veriﬁcation, Journal
of Artiﬁcial Intelligence Research, Vol. 18, pp.83-116, 2003.
 Grunwald, P. D. and Halpern, J. Y., Updating probabilities, Vol. 19, pp.243-278, 2003.
 Guestrin, C., Koller, D., Parr, R. and Venkataraman, S., Eﬃcient solution algorithms for factored
MDPs, Journal of Artiﬁcial Intelligence Research, Vol. 19, pp.399-468, 2003.
 Halal, W. E., Artiﬁcial intelligence is almost here, On the Horizon - The Strategic Planning Resource
for Education Professionals, Vol. 11, No. 2., 2003.
 Halpern, J. Y., Cox’s theorem revisited (technical addendum), Journal of Artiﬁcial Intelligence Re-
search, Vol.11, pp.429-435, 1999.
 Halpern, J. Y., Axiomatising causal reasoning. Journal of Artiﬁcial Intelligence Research, Vol. 12,
 Halpern, J. Y., Conditional plausibility measures and Bayesian networks, Journal of Artiﬁcial Intel-
ligence Research, Vol. 14, pp.359-389, 2001.
 Halpern, J. Y. and Pucella, R., A logic for reasoning about upper probabilities, Journal of Artiﬁcial
Intelligence Research, Vol. 17, pp.s 57-81, 2002.
 Hauskrecht, M., Value-function approximations for partially observable markov decision processes,
Journal of Artiﬁcial Intelligence Research, Vol. 13, pp.33-94, 2000.
 Haynes, R., From alchemy to artiﬁcial intelligence: stereotypes of the scientist in western literature.
Public Understanding of Science, Vol. 12, No. 3, pp.243-253, 2003.
 Hoﬀmann, J., The metric-FF planning system: Translating ’Ignoring delete lists‘ to numeric state
variables, Journal of Artiﬁcial Intelligence Research, Vol. 20, pp.291-341, 2003.
 Hoﬀmann, J. and Nebel, B., The FF planning system: Fast plan generation through heuristic search,
Journal of Artiﬁcial Intelligence Research, Vol. 14, pp.253-302, 2001.
 Holte, R. C. and Choueiry, B. Y., Abstraction and reformulation in artiﬁcial intelligence. Philosoph-
ical Transactions: Biological Sciences, Vol. 358, No. 1435, pp.1197-1204, 2003.
 Hong, J., Goal recognition through goal graph analysis, Journal of Artiﬁcial Intelligence Research,
Vol. 15, pp.1-30, 2001.
“M19N41” — 2008/11/5 — 17:14 — page 567 — #33 i
A Literature Review on Artiﬁcial Intelligence 567
 Howe, A. E. and Dahlman, E., A critical assessment of benchmark comparison in planning, Journal
of Artiﬁcial Intelligence Research, Vol. 17, pp.1-33, 2002.
 Jensen, R. M. and Veloso, M. M., OBDD-based universal planning for synchronised agents in non-
deterministic domains, Journal of Artiﬁcial Intelligence Research, Vol. 13, pp.189-226, 2000.
 Kambhampati, S., Planning graph as a (dynamic)CSP: exploiting EBL, DDB and other CSP search
techniques in Graphplan, Journal of Artiﬁcial Intelligence Research, Vol. 12, pp.1-34, 2000.
 Kaminka, G. A., Pynadath, D. V. and Tambe, M., Monitoring teams by overhearing: A multi-agent
plan-recognition approach, Journal of Artiﬁcial Intelligence Research, Vol. 17, pp.83-135, 2002.
 Kaminka, G. A. and Tambe, M., Robust agent teams via social ly-attentive monitoring, Journal of
Artiﬁcial Intelligence Research, Vol. 12, pp.105-147, 2000.
 Kim, S. M., Irrelevance and relevance of Godel’s theorems to artiﬁcial intelligence. Kybernetes, Vol.
24, No. 4., 1995.
 Kim, S. W. and Kim, S. M., Turing-computability and artiﬁcial intelligence: Godel’s incompleteness
results. Kybernetes, Vol. 24 No. 6., 1995.
 Koehler, J. and Hoﬀmann, J., On reasonable and forced goal orderings and their use in an agenda-
driven planning algorithm, Journal of Artiﬁcial Intelligence Research, Vol. 12, pp.338-386, 2000.
 Kusters, R. and Borgida, A., What’s in an attribute? consequences for the least common subsumer,
Journal of Artiﬁcial Intelligence Research, Vol. 14, pp.167-203, 2001.
 Kvarnstrom, J. and Magnusson, M., TALplanner in IPC-2002: Extensions and control rules, Journal
of Artiﬁcial Intelligence Research, Vol. 20, pp.343-377, 2003.
 Lang, J., Liberatore, P. and Marquis, P., Propositional independence - formula-variable independence
and forgetting, Journal of Artiﬁcial Intelligence Research, Vol. 18, pp.391-443, 2003.
 Leisink, M. and Kappen, B., Bound propagation, Journal of Artiﬁcial Intelligence Research, Vol. 19,
 Lerman, K., Minton, S. N. and Knoblock, C.A., Wrapper maintenance: A machine learning approach,
Journal of Artiﬁcial Intelligence Research, Vol. 18, pp.149-181, 2003.
 Lin, F., Compiling causal theories to successor state axioms and STRIPS-like systems, Journal of
Artiﬁcial Intelligence Research, Vol. 19, pp.279-314, 2003.
 Long, D. and Fox, M., The 3rd international planning competition: Results and analysis, Journal of
Artiﬁcial Intelligence Research, Vol. 20, pp.1-59, 2003.
 Lusena, C., Goldsmith, J. and Mundhenk, M., Nonapproximability results for partially observable
markov decision processes, Journal of Artiﬁcial Intelligence Research, Vol. 14, pp.83-103, 2001.
 Markham, I. S., Mathieu, R. G. and Wray, B. A., Kanban setting through artiﬁcial intelligence: a
comparative study of artiﬁcial neural networks and decision trees, Integrated Manufacturing Systems,
Vol. 11, No. 4., 2000.
 Masnikosa, V. P., The fundamental problem of an artiﬁcial intelligence realisation, Kybernetes, Vol.
27, No. 1., 1998.
 Maynard-Zhang, P. and Lehmann, D., Representing and aggregating conﬂicting beliefs, Journal of
Artiﬁcial Intelligence Research, Vol. 19, pp.155-203, 2003.
 McDermott, D., PDDL2.1 - The art of the possible? Commentary on Fox and Long. Journal of
Artiﬁcial Intelligence Research, Vol. 20, pp.145-148, 2003.
 Meek, C., Finding a path is harder than ﬁnding a tree, Journal of Artiﬁcial Intelligence Research,
Vol. 15, pp.383-389, 2001.
 Metaxiotis, K., Ergazakis, K., Samouilidis, E. and Psarras, J., Decision support through knowledge
management: the role of the artiﬁcial intelligence. Information Management & Computer Security;
Vol. 11 No. 5., 2003.
 Moraga, C., Trillas, E. and Guadarrama, S., Multiple-valued logic and artiﬁcial intelligence fun-
damentals of fuzzy control revisited. Artiﬁcial Intelligence Review, Vol. 20, No. 3-4, pp.169-197,
“M19N41” — 2008/11/5 — 17:14 — page 568 — #34 i
568 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
 Nau, D. S., Au, T. C., Ilghami, O., Kuter, U., Murdock, J. W., Wu, D. and Yaman, F., SHOP2: An
HTN Planning System, Vol. 20, pp.379-404, 2003.
 Neal, R. M., On deducing conditional independence from d-separation in causal graphs with feedback
(Research Note), Journal of Artiﬁcial Intelligence Research, Vol. 12, pp.87-91, 2000.
 Nebel, B., On the compilability and expressive power of propositional planning formalisms, Journal
of Artiﬁcial Intelligence Research, Vol. 12, pp.271-315, 2000.
 Nock, R., Inducing interpretable voting classiﬁers without trading accuracy for simplicity: theoretical
results, approximation algorithms, and experiments, Journal of Artiﬁcial Intelligence Research, Vol.
17, pp.137-170, 2002.
 Opitz, D., Maclin, R., Popular ensemble methods: an empirical study, Journal of Artiﬁcial Intelli-
gence Research, Vol. 11, pp.169-198, 1999.
 Palomar, M. and Martinez-Barco, P., Computational approach to anaphora resolution in spanish
dialogues, Journal of Artiﬁcial Intelligence Research, Vol. 15, pp.263-287, 2001.
 Patel-Schneider, P. F. and Sebastiani, R., A new general method to generate random modal formulae
for testing decision procedures, Journal of Artiﬁcial Intelligence Research, Vol. 18, pp.351-389, 2003.
 Peng Y. and Zhang X., Integrative data mining in systems biology: from text to network mining,
Artiﬁcial Intelligence in Medicine, Vol. 41, No. 2, pp.83-86, 2007,
 Peral, J., Ferrandez, A., Translation of pronominal anaphora between English and spanish: Discrep-
ancies and evaluation, Journal of Artiﬁcial Intelligence Research, Vol. 18, pp. 117-147, 2003.
 Plenert, G., Improved decision support systems help to build better artiﬁcial intelligence systems.
Kybernetes, Vol. 23 No. 9., 1994.
 Poole, D. and Zhang, N.L., Exploiting contextual independence in probabilistic inference, Journal of
Artiﬁcial Intelligence Research, Vol. 18, pp.263-313, 2003.
 Price, B. and Boutilier, C., Accelerating reinforcement learning through implicit imitation, Journal
of Artiﬁcial Intelligence Research, Vol. 19, pp.569-629, 2003.
 Pynadath, D.V., Tambe, M., The communicative multi-agent team decision problem: Analysing
teamwork theories and models, Journal of Artiﬁcial Intelligence Research, Vol. 16, pp. 389-423,
 Raynor, W. J., The international dictionary of artiﬁcial intelligence. Reference Reviews, Vol. 14
No. 6., 2000.
 Refanidis, I. and Vlahavas, I., The GRT planning system: backward heuristic construction in forward
state-space planning, Journal of Artiﬁcial Intelligence Research, Vol. 15, pp.115-161, 2001.
 Reiter, E., Sripada, S. G. and Robertson, R., Acquiring correct knowledge for natural language
generation, Journal of Artiﬁcial Intelligence Research, Vol. 18, pp.491-516, 2003.
 Renz, J. and Nebel, B., Eﬃcient methods for qualitative spatial reasoning, Journal of Artiﬁcial
Intelligence Research, Vol. 15, pp.s 289-318, 2001.
 Resnik, P., Semantic similarity in a taxonomy: an information-based measure and its application
to problems of ambiguity in natural language, Journal of Artiﬁcial Intelligence Research, Vol. 11,
 Rosati, R., Reasoning about minimal belief and negation as failure, Journal of Artiﬁcial Intelligence
Research, Vol. 11, pp.277-300, 1999.
 Sanchez, R. and Kambhampati, S., AltAltp: Online parallelization of plans with heuristic state search,
Journal of Artiﬁcial Intelligence Research, Vol. 19, pp.s 631-657, 2003.
 Sato, T. and Kameya, Y., Parameter learning of logic programs for symbolic-statistical modeling,
Journal of Artiﬁcial Intelligence Research, Vol. 15, pp.391-454, 2001.
 Scerri, P., Pynadath, D. V. and Tambe, M., Towards adjustable autonomy for the real world, Journal
of Artiﬁcial Intelligence Research, Vol. 17, pp.171-228, 2002.
“M19N41” — 2008/11/5 — 17:14 — page 569 — #35 i
A Literature Review on Artiﬁcial Intelligence 569
 Schiaﬀonati, V., A framework for the foundation of the philosophy of artiﬁcial intelligence. Minds
and Machines, Vol. 13, No. 4, pp.537-552, 2003.
 Schlimmer, J. C. and Hermens, L.A., Software agents: completing patterns and constructing user
interfaces, Journal of Artiﬁcial Intelligence Research, Vol. 1, pp.61-89, 1993.
 Shatkay, H. and Kaelbling, L. P., Learning geometrical ly-constrained hidden markov models for robot
navigation: Bridging the topological-geometrical gap, Journal of Artiﬁcial Intelligence Research, Vol.
16, pp.167-207, 2002.
 Singer, J., Gent, I. P. and Smaill, A., Backbone fragility and the local search cost peak, Journal of
Artiﬁcial Intelligence Research, Vol. 12, pp.235-270, 2000.
 Singh, S., Litman, D., Kearns, M. and Walker, M., Optimising dialogue management with reinforce-
ment learning: experiments with the NJFun system, Journal of Artiﬁcial Intelligence Research, Vol.
16, pp.105-133, 2002.
 Siskind, J. M., Grounding the lexical semantics of verbs in visual perception using force dynamics
and event logic, Journal of Artiﬁcial Intelligence Research, Vol. 15, pp.31-90, 2001.
 Sivramkrishna, S. and Panigrahi R., Articulating uneven regional development: artiﬁcial intelligence
as a tool in development planning. Journal of Human Development, Vol. 4, No. 3, pp.437-456, 2003.
 Smith, D. E., The case for durative actions: A commentary on PDDL2.1, Journal of Artiﬁcial
Intelligence Research, Vol. 20, pp.149-154, 2003.
 Soderland, S., Lehnert, W., Wrap-up: a trainable discourse module for information extraction, Jour-
nal of Artiﬁcial Intelligence Research, Vol. 2, pp.131-158, 1994.
 Stefanuk, V. L. and Zhozhikashvili, A. V., Productions and rules in artiﬁcial intelligence, Kybernetes:
The International Journal of Systems & Cybernetics, Vol. 31 No. 6., 2002.
 Stone, P., Littman, M.L., Singh, S., Kearns, M., ATTAC-2000: An adaptive autonomous bidding
agent, Journal of Artiﬁcial Intelligence Research, Vol. 15, pp. 189-206, 2001.
 Stone, P., Schapire, R. E., Littman, M. L., Csirik, J. A. and McAllester, D., Decision-theoretic
bidding based on learned density models in simultaneous, interacting auctions, Journal of Artiﬁcial
Intelligence Research, Vol. 19, pp.209-242, 2003.
 Straccia, U., Reasoning within fuzzy description logics, Journal of Artiﬁcial Intelligence Research,
Vol. 14, pp.137-166, 2001.
 Tan, K. C., Khor, E. F., Lee, T. H. and Sathikannan, R., An evolutionary algorithm with advanced
goal and priority speciﬁcation for multi-objective optimisation, Journal of Artiﬁcial Intelligence Re-
search, Vol. 18, pp.183-215, 2003.
 Tay, D. P. H. and Ho, D. K. H., Artiﬁcial intelligence and the mass appraisal of residential apart-
ments, Journal of Property Valuation and Investment, Vol. 10 No. 2., 1992.
 Tennenholtz, M., Competitive safety analysis: Robust decision-making in multi-agent systems, Jour-
nal of Artiﬁcial Intelligence Research, Vol. 17, pp.363-378, 2002.
 Thompson, C. A. and Mooney, R. J., Acquiring word-meaning mappings for natural language inter-
faces, Journal of Artiﬁcial Intelligence Research, Vol. 18, pp.1-44, 2003.
 Tobies, S., The complexity of reasoning with cardinality restrictions and nominals in expressive
description logics, Journal of Artiﬁcial Intelligence Research, Vol. 12, pp.199-217, 2000.
 Toni, A. D., Nassmbeni, G. and Tonchia, S., An artiﬁcial, intelligence-based production scheduler,
Integrated Manufacturing Systems, Vol. 7 No. 3., 1996.
 Turney, P. D., Cost-sensitive classiﬁcation: empirical evaluation of a hybrid genetic decision tree
induction algorithm , Journal of Artiﬁcial Intelligence Research, Vol. 2, pp.369-409, 1995.
 Walker, M. A., An application of reinforcement learning to dialogue strategy selection in a spoken
dialogue system for email, Journal of Artiﬁcial Intelligence Research, Vol. 12, pp.387-416, 2000.
 Walker, M.A., Langkilde-Geary, I., Wright Hastie, H., Wright, J., Gorin, A., Automatical ly train-
ing a problematic dialogue predictor for a spoken dialogue system, Journal of Artiﬁcial Intelligence
Research, Vol. 16, pp. 293-319, 2002.
“M19N41” — 2008/11/5 — 17:14 — page 570 — #36 i
570 International Journal of Information and Management Sciences, Vol. 19, No. 4, December, 2008
 Walsh, W. E. and Wellman, M. P., Decentralised supply chain formation: A market protocol and
competitive equilibrium analysis, Journal of Artiﬁcial Intelligence Research, Vol. 19, pp.513-567,
 Wang S., Wang Y., Du W., Sun F., Wang X., Zhou C. and Liang Y., A multi-approaches-guided
genetic algorithm with application to operon prediction, Artiﬁcial Intelligence in Medicine, Vol. 41,
No. 2, pp.151-159, 2007.
 Weiss, G. M. and Provost, F., Learning when training data are costly: The eﬀect of class distribution
on tree induction, Journal of Artiﬁcial Intelligence Research, Vol. 19, pp.315-354, 2003.
 Wiewiora, E., Potential-based shaping and Q-value initialisation are equivalent, Journal of Artiﬁcial
Intelligence Research, Vol. 19, pp.205-208, 2003.
 Wilkins, D. E., Lee, T. J. and Berry, P., Interactive execution monitoring of agent teams, Journal of
Artiﬁcial Intelligence Research, Vol. 18, pp.217-261, 2003.
 Wolpert, D. H. and Tumer, K., Collective intel ligence, data routing and braess’ paradox, Journal of
Artiﬁcial Intelligence Research, Vol. 16, pp.359-387, 2002.
 Wongpinunwatana, N., Ferguson, C. and Bowen, P., An experimental investigation of the eﬀects of
artiﬁcial intelligence systems on the training of novice auditors. Managerial Auditing Journal, Vol.
15 No. 6., 2000.
 Wray, R. E. and Laird, J.E., An architectural approach to ensuring consistency in hierarchical exe-
cution, Journal of Artiﬁcial Intelligence Research, Vol. 19, pp.355-398, 2003.
 Xu, K. and Li, W., Exact phase transitions in random constraint satisfaction problems, Journal of
Artiﬁcial Intelligence Research, Vol. 12, pp.93-103, 2000.
 Xu, X., He, H. and Hu, D., Eﬃcient reinforcement learning using recursive least-squares methods,
Journal of Artiﬁcial Intelligence Research, Vol. 16, pp.s 259-292, 2002.
 Younes, H. L. S. and Simmons, R. G., VHPOP: Versatile heuristic partial order planner, Journal of
Artiﬁcial Intelligence Research, Vol. 20, pp.405-430, 2003.
 Ygge, F. and Akkermans, H., Decentralised markets versus central control: a comparative study,
Journal of Artiﬁcial Intelligence Research, Vol. 11, pp.301-333, 1999.
 Zanuttini, B., New polynomial classes for logic-based abduction, Journal of Artiﬁcial Intelligence
Research, Vol. 19, pp.1-10, 2003.
 Zhang, N. L. and Zhang, W., Speeding up the convergence of value iteration in partially observable
markov decision processes, Journal of Artiﬁcial Intelligence Research, Vol. 14, pp.29-51, 2001.
 Zhou X., Liu B., Wu Z. and Feng Y., Integrative mining of traditional Chines medicine literature
and MEDLINE for functional gene networks, Artiﬁcial Intelligence in Medicine, Vol. 41, No. 2,
 Zucker, J. D., A grounded theory of abstraction in artiﬁcial intelligence, philosophical transactions:
biological sciences, Journal of Artiﬁcial Intelligence Research, Vol. 358 No. 1435, pp.1293-1309,
S.A. Oke is currently a lecturer in the Department of Mechanical Engineering, University of Lagos,
Nigeria. He received hid B.Sc. (Hons) and M.Sc. in Industrial Engineering from the University of
Ibadan, Nigeria and currently a doctoral candidate in the same department. His research interests are
industrial engineering and soft computing methodologies.
Department of Mechanical Engineering, University of Lagos, Lagos, Nigeria.
E-mail: sa firstname.lastname@example.org