Pergamon

I t l t . Ttut1.s. Opl R r s . Vol. 1 , No. 2 . pp. 169 1x7. 1994 Elsevier Science Ltd. Copyright t', 1994 IFORS

Printed in Great Britain. All rights reserved 09694016(93)E0018-8 0969-601 6/94 $6.00 + 0.00

Towards Intelligent Organisational Information Systems

GREGORY MENTZAS National Technical University of Athens, Greece

I t has been argued in the literature that although specific types of computer-based information systems (CBIS) are powerful tools in certain parts of the decision-making processes in modern organisations, none of them provides integrated support. This lack becomes even more acute when one considers the operation of CBIS within an organisational setting where consistent assistance of a multitude of users in different departments is required, together with facilities for modelling such features as cooperation, conflict, negotiation, etc. On the other hand, there are indications that the incorporation of abilities such as perception, interpretation, reasoning, explanation, goal-setting, and learning in computer-aided support would greatly enhance the aforementioned decision-making processes. This paper reviews the evolution of CBIS, and attempts to synthesise current research towards the goal of competent and intelligent aiding of decision-making in organisational settings. The main argument of the paper is that decision-making functions in multi-participant organisations can be facilitated by the use of intelligent software entities with autonomous processing capabilities which possess coordination and negotiation facilities and are organised in distributed, hierarchical societies. This paper presents a conceptual definition of these entities, outlines their structural characteristics, and describes a framework for research towards their development.

Key words: information systems, organisational environment, decision-making, artificial intelligence

INTRODUCTION

Various studies have examined the relationship of research in computer-based information systems (CBIS), in comparison with operations research (OR) and management science (MS) methods, techniques and practice. It has been found that although specific CBIS types (e.g. decision support systems, expert systems, etc.) are powerful tools in certain parts of the decision processes, none of the existing types provides adequate support for all the major decision-making functions, and none covers in an adequate manner all information needs in the modern organisation [see e.g. Parker and Al-Utaibi (1 986) for a discussion of the shortcomings inherent in management information systems, Turban and Trippi (1989) and Partridge ( 1 987) for analyses of expert systems limitations].

The famous decision-making model proposed by Simon (1977) has been repeatedly used for illustrating this lack of complete support. Simon characterised the decision-making process in terms of three main issues:

intelligence: searching the environment for opportunities calling for a decision; design: inventing, developing and analysing possible courses of action; and choice: selecting a particular course of action from those available.

Parker and Al-Utaibi (1986) argue that while management information systems (MIS), for example, have made considerable progress in the intelligence phase, decision support systems assist in the design choice phase, and OR/MS techniques facilitate the choice phase. On the other hand, Turban and Trippi (1989) argue that there are many opportunities for cooperation and cross-fertilisation between OR and expert systems (p. 319). Hence, various techniques have been proposed for the integration of approaches, so that one covers the inefficiencies of the other [see e.g. Turban and Watkins (1986) for research aiming at the integration of expert systems and decision support systems].

The above shortcomings of existing CBIS become even more acute when one considers their operation within an organisational setting. In such an environment, the aim is no longer the support of a single user, but it calls for the consistent assistance of a multitude of users in different

Correspondence: Dr Greyory Mentzus. Depurtment of Electricul und Computer Enqinrrrimg. Nutiotltrl Tc.chictrl U t I iwrs i t~~ o f ' Athens. 4 2 . 28th October str.. 10682 Athens. Greece. Tel. 301-3616924, fax 301-3626792, e-mail: gmcntzas.thescns.tltu~l.gr

169

I 70 Gregory Mentzas-Intelligent Organisational Information Systems

departments, with varying needs and requires a coherent framework for modelling such features as cooperation, conflict, negotiation, etc. (see e.g. Applegate, 1991). Furthermore, there is a need to support the distributed computing framework, within which modern organisations operate [e.g. Quarterman and Carl-Mitchell (1993) discuss the shift from centralised to distributed computing].

The starting point of the present paper is the argument of Turban and Trippi (1989, p. 319), that a fruitful area of cooperation between expert systems and operations research is that of the ‘simulation of systems possessing abilities such as perception, interpretation, reasoning, explanation, goal- setting, and learning’. In addition, this paper explores the thesis that efficient organisational support should incorporate features that would simulate the various types of human interaction, so that crucial organisational features, such as parallel work and geographical distribution of operations, can be effectively supported.

This paper attempts to trace the evolution of computer-aided information and decision support and study the possible role of current research. Specifically, the paper argues that recent advances in object-orientation and distributed artificial intelligence open up the prospect of techniques that will enable the aforementioned cross-fertilisation and lead to the so-called intelligent organisational information systems.

This paper classifies the various types of computer-based information systems (CBIS) that have been utilised for information and decision support. Seven categories are examined; management information systems; decision support systems; group decision support systems; expert systems; office information systems; executive information systems; and intelligent organisational informa- tion systems. The latter category constitutes an attempt to synthesise research in such fields as computer science, operations research and artificial intelligence towards the common goal of competent and intelligent aiding in organisational settings. In a study of the constituent elements of the various types of CBIS, we examine five such elements (data, models, knowledge, cooperation, and support for man-machine interaction), and review their evolution.

Then, we elaborate on the issue of efficient and intelligent assistance in formation and decision-making support. We argue that enterprise-wide information handling and decision support could be greatly facilitated by the existence of intelligent software entities with autonomous processing capabilities which possess coordination and negotiation facilities and are organised in distributed, hierarchical societies. The paper presents a conceptual definition of such entities, outlines their structural characteristics, and describes a framework for research towards the development and population of their societies, i.e. intelligent organisational information systems.

Note that this paper deals only with computer-based information systems (CBIS) and does not analyse the work done concerning the sociorganisational aspects of these systems (see e.g. Zennetos and Sertel, 1970; Matsuda, 1993), or the interdependencies between information systems and organisational structure (see e.g. Markus and Robey, 1988; Huber, 1982).

This paper is structured in the following way. The next section gives a review of the evolution of computer-based information systems and their main characteristics. We proceed by examining the constituent elements of CBIS. The fourth section outlines the basic architectural and structural issues related to the design and implementation of active intelligence assistance, while the last section summarises the findings and identifies directions for future research.

COMPUTER-BASED INFORMATION SYSTEMS

Classification of systems

Several computer-based systems have been developed to deliver the technology necessary to provide support for decision making and information management. They can be classified in seven types of systems:

management information systems (MIS); decision support systems (DSS); group decision support systems (GDSS); expert systems (ES);

Internutionul Trunsuctions in Operutionul Research Vol. l , No. 2 171

ofice information systems (01s); executive information systems (EIS); and intelligent organisational information systems (101s).

Table l summarises the major types of these systems and their role in decision support.

Table I . Types of computer-based information systems

Type of system Role

Management information systems (MIS)

Decision support systems

Group decision support systems (GDSS) Expert systems (ES)

Office information systems (01s)

Executive information systems (EIS)

Intelligent organisational information systems (101s)

Analysis of information, generation of requested reports, solving of structured problems

Use of data, models and decision aids in the analysis of semi-structured problems for individuals

Extension of DSS with negotiation and communication facilities for groups

Capturing and organisation of corporate knowledge about an application domain and translation into expert advice Support of the office worker in the effective and timely management of office objects, the goal-oriented and ill-defined office processes and the control of information flow in the office

Evaluation of information in timely analyses for top-level managerial levels, in an intelligent manner

Assistance (and independent action) in all phases of decision making and information support in multi-participant organisations.

Management information systems

The literature abounds with accounts of what is, or is not a management information system (MIS) (see also Parker and Al-Utaibi, 1986): the definitions range from the extreme of those who believe that an MIS is a computer-based system that produces an expanded set of reports and has a query capability, to those claiming that an MIS serves all the information needs of an organisation.

What constitutes managerial decision-making? According to Ackoff (1967) managerial decisions can be classified into three types:

decisions for which adequate models are available or can be constructed and from which optimal

decisions for which adequate models can be constructed but from which optimal solutions cannot

decisions for which adequate models cannot be constructed (unstructured decision-making).

Keen and Scott-Morton (1 978) argue that the main area of impact of MIS on organisations refers to the structured tasks where standard operating procedures, decision rules and information flows can be reliably predefined. They argue that MIS improve efficiency by reducing costs and repiacing clerical personnel, while the relevance of MIS to decision-making is limited to providing reports and access to data.

Hence, it can be considered that the role played by MIS refers only to the first level of Ackoff’s classification, i.e. MIS have failed to support semi-structured and even unstructured decision- making (see also Watson and Hill, 1983). The use of computer-based systems that would support and enhance the decision-maker’s judgement in areas of unstructured and semi-structured decision- making is where decision support systems come in.

(or near optimal) solutions can be derived (structured decision-making)

be extracted (semi-structured decision-making)

Decision support systems

As Keen (1987) has put it, ‘. . . right from the start of the DSS movement, and even now, there has been no established definition of DSS . . .’. For example, according to Sprague (1980) DSS are

I72 Gregory Mentzas-Intelligent Organisational Information Systems

‘interactive computer-based systems, which help decision makers utilise data and models to solve unstructured problems’. An alternative definition of decision support systems (DSS) is given in Edwards (1992): a decision support system is a ‘system which enables the user to access data and/or models so that he or she may take better decisions’. This definition is neutral with respect to issues that are discussed extensively within the DSS field, e.g. whether a DSS must be computer-based, whether it must include a normative model, etc. (see e.g. Sprague and Carlson, 1982; Bonczek et al., 1981).

The issue of incorporation of normative models has been examined in the empirical work on the use of DSS that was carried out by Alter (1980). The classification provided by Alter, according to the generic operations provided by the specific DSS, distinguishes seven types: data-analysis systems; analysis information systems; accounting models; representational models; optimisation models; and suggestion models.

The DSS literature has seen a wealth of research efforts; hundreds of DSS have been constructed, in order to facilitate decision-making in a variety of situations [see e.g. Eom and Lee (1990) for a survey of applications].

Group decision support systems

It has been argued that group activities are economically necessary, efficient as means of production and reinforcing of democratic values (see e.g. Kraemer and King, 1988; Hatcher, 1992). DeSanctis and Gallupe (1987) provide the definition of a GDSS as ‘an interactive computer-based system that facilitates the solution of unstructured problems by a set of decision-makers working together as a group’. Operationally this means increasing the speed at which decisions are reached without reducing, and hopefully enhancing, the quality of resulting decisions. Shaw (1981 ), for example, concluded that groups produce more and better solutions to problems than do individuals, particularly on judgmental tasks.

The primary problems of productivity loss in group decision meetings are from information loss, information distortion, or suboptimal decision making (i.e. not enough issues and alternatives are explored) (see Kraemer and King, 1988). These authors conceive the GDSS as a sociotechnical ‘package’ comprised of:

Hardware: Hardware includes the conference facilities and the computing, telecommunications and audiovisual equipment. Software: Software includes database management systems, high level programming languages, and decision modelling and support software. Decision modelling includes software specifically tailored for group analysis and decision support includes modelling languages, decision structuring techniques (such as brainstorming, Delphi technique, etc.), utility and probability assessment techniques, multiattribute analysis, etc. Organisationware: Organisationware includes the organisational data, group processes and management procedures for collaborative group work. People: People refers to the participants in the group and the support staff.

There are two main research streams in GDSS. The first has concentrated on discovering the psychological or cognitive processes of individuals and groups involved in reaching conclusions and on the sociology of small-group interactions. The other major stream has been the development of technologically supported means of collecting, managing, displaying and coordinating information that might be useful in decision situations (see e.g. Vogel and Nunamaker, 1988).

Expert systems

Expert systems (ES) have been defined as systems that embody knowledge, offer intelligent advice or take intelligent decisions about a processing function [see e.g. Edwards (1 992) for some definitions]. The major issue in expert systems is that they ‘replace’ the human expert, by embodying his/her expertise within an electronic expert.

hternutionul Transactions in Operational Research Vol. I , No. 2 173

It has been argued that expert system development is quite different from that of conventional systems; the reasons for this difference can be summarised in the following: conventional programs deal with problems that have been solved beforehand, hence they try to change knowledge, procedures or algorithms into code in an efficient manner, while ES determine and encode expert knowledge, based upon knowledge acquisition with repeated interviews of human experts (see Williams, 1986).

The early introduction of expert systems was accompanied with the use of artificial intelligence techniques that refer to knowledge acquisition, knowledge structuring, inference mechanisms, search strategies, symbolic representation and truth maintenance.

Early ES were focused on modular knowledge bases in conjunction with efficient logical inference strategies. Partridge (1 987) claimed that the major assumptions upon which expert systems rest are: the fact that necessary knowledge can be represented as a collection of more or less independent rules; and that intelligent decision making can be implemented as a logical, truth-derivation mechanism. He argues that these assumptions are true in domains of abstract, technical expertise, such as mathematics, geology, chemistry, configuring computer systems, medical diagnosis, game playing and puzzle solving. However, these assumptions are weak in the following domains: natural language processing; intelligent tutoring; self-explanation of behaviour; and advanced robots. In addition, knowledge in ES is static; learning, adaptation and uncertainty mechanisms are not included.

Ofice information systems

Office information systems (01s) aim at supporting the document-related, procedural and communication issues of office work (see Ellis and Nutt, 1980). They have been modelled as encompassing three domains: passive office objects; office procedures; and office tasks (see Mentzas, 1991).

Office objects are the primitive office elements; examples of office objects are documents, files, printers, etc.; hence, office objects provide metaphors that represent their actual counterparts in the physical office. Office procedures can be considered a set of mappings among office objects; office procedures are routine sequences of operations that are used to manipulate office objects. They model the event-driven behaviour of office work and are triggered upon completion of some awaited event, e.g. the arrival of a message, the completion of a form, or the modification of a document. Finally, office tasks are goal-directed and cannot necessarily be encoded to a precise procedure to be followed. Their intention is to model cooperation among many office agents, negotiation among parties, confrontation and argumentation, and the abilities to learn and reach goals.

Two important requirements of 01s with reference to office objects concern the need to support concurrent sharing (e.g. the fact that several office workers from different departments may wish to examine the same information at the same time) and referential sharing, e.g. a purchase order might be replicated within the purchasing department as well as the department from which it originated; nevertheless, the status of the order should be maintained consistently in both departments.

Office procedures model dynamically changing work-flow within an office; they focus on the handling of information flow within corporate processing involved in performing a particular process. Office procedures need to be executed in a parallel manner, so that concurrency of office work is achieved, and they need to be stored and retrieved in a ‘dynamic’ manner, so that queries of the present state of a procedure would give a persistent picture. In addition, message-passing protocols between office procedures guarantee the modelling of work interdependencies. Finally, office tasks, being knowledge-intensive, since they represent the corporate level rules and procedures, lend themselves to artificial intelligence techniques.

Executive information systems

While DSS attempt to answer specified management questions, there is a growing need for timely information needed by senior managers, for intelligent questions about particular aspects of an

I74 Greyory Mentzas-Intelligent Organisational Information Systems

organisation. Executive information systems (EIS, or executive support systems, EPS) have been created to monitor the decision environment, evaluate the captured information for opportunities and/or problems, and present timely analyses to top-level managers. An EIS can be defined as ‘a computerized system that provides executives with easy access to internal and external information that is relevant to their critical success factors’ (see Watson et al., 1991).

In a sense, EIS strive to accomplish office automation facilities (like OIS), decision analysis support (like DSS) and intelligence (like ES) (see e.g. Rockart and DeLong, 1988; Turban, 1990).

Towards intelligent organisational injormation systems

There are two main problems facing all of the above types of systems. First, none of them satisfies in an adequate manner all the decision- and information-related processes of an organisation. On the other hand, it is clear that organisational productivity can be maximised by creating, using and maintaining structural and dynamic configurations of multi-participant interaction activities; hence, supporting the latter activities is of high importance for computing technology (see Applegate et al., 1991). The second major problem of existing systems is that none of them covers explicitly the needs of large-scale organisations, e.g. support of parallel work; intelligent assistance in group communica- tion; negotiation and conflict; distribution of processing and reasoning facilities; techniques for multi-participant planning; organisational learning facilities; etc. Hence, there is clearly a need for intelligent support systems that would facilitate decision-making in the organisational environment.

With the term intelligent organisational information systems (101s) we label systems of intelligent software entities which are organised in loosely coupled, distributed architectures, and include communication, control, and task-sharing facilities, with the addition of effectuation and advanced modelling capabilities. Intelligent organisational information systems aim to relieve the burdens and assumptions imposed with ES and EIS. Actually, while ES impose a centralised form of coordination, through the use of (usually) blackboard-based architectures, 101s depend on the basic features of distributed artificial intelligence, coupled with object-oriented programming ideas. We claim that recent advances in the fields of distributed artificial intelligence and object-oriented computing facilitate the design and development of 101s. In the following we review the basic technological background of IOTS, i.e. distributed artificial intelligence and object-oriented computation.

Technological background

Distributed artificial intelligence (DAI) proposes a very different approach to the design and construction of intelligent systems, to that advocated by ES. I t proposes the provision of intelligence via a federation of co-operating intelligent ‘agents’. The characteristic of each agent is that it possesses expertise in a particular area, or that it provides the capability to effect a particular function. Bond and Gasser (1988) divide work into DAI in two primary areas: distributed problem solving (DPS) and multi-agent systems (MAS); a third area, parallel AI, will not be considered here. DPS considers how the work of solving a particular problem can be divided between a number of processing ‘nodes’, while MAS research is concerned with coordinating intelligent behaviour between a collection of (possibly pre-existing) autonomous intelligent ‘agents’, which can coordinate their knowledge, goals, skills, and plans jointly to take action or to solve problems. The agents in a MAS may be working towards a single global goal, or towards separate individual goals that interact. MAS agents must share knowledge about problems and solutions. In addition, they must reason about the processes of coordination among the agents. No firm definition of an ‘agent’ is given by Bond and Gasser (1988). They state that they rely on a simple and intuitive notion of an agent as a computational process with a single locus of control and/or ‘intention’.

On the other hand, object-oriented computation in examining concurrency in programming languages further enhances the original intent of the Simula language to describe simulations of real systems, since the real world is concurrent and distributed. One of the earliest formalisms for concurrent object-based programming is based on the actor formalism proposed by Hewitt (1977).

Intcv-nutionul Trunsuctions in Operutionul Reseurch Vol. I , N o . 2 175

Hewitt’s model uses message-passing between actors to represent control structures such as request/reply and recursion (see also Agha, 1990). The message-passing metaphor treats objects as autonomous entities that synchronise and exchange information with one another only by explicitly sending messages. This view has been considered equivalent to stating that each object ‘encapsulates’ some local state that may be accessed only by methods that are somehow associated with the object; objects may access the local state of another object only by requesting that the recipient of a message execute some method.

Various architectures have been proposed in the literature that can provide the initial framework for the development of 101s [see e.g. Papazoglou et al. (1992), in the case of intelligent cooperative database systems, O’Hare (1 990) in manufacturing systems, Mentzas (1 993b) in the area of office and production automation, and Rose et al. (1992) in CASE applications]. Such architectures draw from findings in the use of object-oriented and distributed artificial intelligence research in manufacturing systems (Hynynen, 1992), office information systems (Mentzas, 1991 ), as well as from studies in cooperation and coordination (see e.g. Durfee, 1988; Ellis et al., 1991 ; Mentzas, 1993a; Ching et al., 1993).

ANALYSIS OF CBIS

Components of computer-based systems

The previous section briefly reviewed the major types of CBIS and their characteristics. Three major types of activities are accomplished with these systems:

information reporting (e.g. with MIS); communication and negotiation activities (e.g. with GDSS); and decision activities (e.g. support of selection of alternatives in DSS and ES).

Although all seven types of computer-based systems try to cover the three aforementioned activities, they exhibit varying degrees of coverage. Table 2 presents the degree to which each type of system attempts to cover these activities. It appears that information management support has reached relatively high scores in the MIS, 01s and EIS types of systems, while decision support has been adequately provided with DSS, GDSS, ES and EIS types of systems. A significant problem with communication and negotiation support can be usually found in the MIS, DSS, and ES, while this necessity is covered in GDSS and 01s.

Table 2. Types of support in computer-based information systems

Information system Information support Decision support Communication support

MIS DSS GDSS ES 01s EIS 101s

High Medium Medium Medium High High High

L0 W

High High High Low High High

Low Low High Low High Low High

The abbreviations of the various types of information systems are as follows: MIS: management information systems; DSS: decision support systems; GDSS: group decision support systems; ES: expert systems; 01s: ofice information systems; EIS: executive information systems; 101s: intelligent organisational information systems.

The CBIS examined here include a number of constituent elements, which in turn are built using fundamental theoretical techniques drawn from such fields as operations research, computer programming, artificial intelligence, etc. We can distinguish among five such constituents: database management systems (DBMS); model management systems (MMS); knowledge management systems (KBMS); cooperative management systems (CMS); and dialogue systems (DS).

I76 Gregory Mentzas-Intelligent Organisational Information Systems

Database management systems provide mechanisms for information storage, retrieval and processing. Model management systems use generalised modelling languages, which, when coupled to a variety of algorithms, are capable of representing and evaluating multiple mathematical abstractions of a given system. Knowledge management systems capture the knowledge of experts in a field and possess inference procedures capable of searching the state-space of available knowledge and solving problems that are difficult enough to require significant human expertise for their solution. Cooperation management systems facilitate conflict resolution and negotiation support when two or more decision makers are involved in the process of decision. Dialogue systems are request and response mechanisms, which support human-machine interac- tions as two-way message passing activities.

Table 3. Elements of computer-based information systems

Information system DBMS MMS KBMS CMS DS

MIS Basic DSS Basic GDSS Basic ES Basic 01s Basic EIS Basic 101s Basic

Basic Basic Optional Basic Basic Optional Basic Basic Optional Basic Basic

Optional Basic Basic Basic Basic Basic Basic Basic Basic

The abbreviations of the various types of information systems are as follows: MIS: management information systems; DSS: decision support systems; GDSS: group decision support systems; ES: expert systems; 01s: office information systems; EIS: executive information systems; 101s: intelligent organisation information systems.

The abbreviations used for the elements of the systems are as follows: DBMS: database management system; MMS: model management system; CMS: cooperation management system; KBMS: knowledge-based management system; DS: dialogue system.

Table 3 attempts an analysis of the computer-based systems for decision support regarding their constituents. Some elements are considered basic for a system, since they provide the fundamental mechanisms for a system to be classified in a specific category, while some others are considered optional, in the sense that although they are not necessary for the system, some implementations found in the literature include them. Table 4 reviews the basic characteristics of CBIS, in terms of their focus, the range of decision problems supported, the level of organisational support provided, the degree of organisational integration they offer, the main goals of the decision-making aid, the types of decision models they may include, and whether they support individuals or groups of decision-makers.

In the following we attempt a closer look at the above mentioned elements, and analyse the possibilities of embedding object-oriented and knowledge-related facilities, with the aim of examining the possible evolution towards intelligent organisational information systems.

Data management

Various types of databases have been used in computer-based systems; any database adhering to the hierarchical, CODASYL network or relational data model can be candidate for inclusion in such a system (see Sprague and Carlson, 1982). Nevertheless, databases designed with the more semanti- cally oriented post-relational data models can provide enhanced storage and retrieval mechanisms. Hence, one may envision the co-existence of multiple databases of differing types within one system and their mapping on to one common framework.

Two major problems arise in this case. The first deals with the provision of a unifying framework for covering the multiplicity of database representations. Predicate calculus techniques have been proposed towards this direction, but their implementation is still immature [see Reiter ( 1 985) for the

International Transactions in Operational R

esearch V

ol. I, No. 2

177

U

c .- U .- a P C

cd

ITOR 1:Z-E

178 Gregory Mentzas-Intelligent Organisational Information Systems

expression of the relational data model in first-order logic and Li (1985) for a recasting of the entity-relationship and the semantic data models in predicate calculus]. The logic-oriented approach, however, has still unanswered questions to deal with; they refer to query definition, incomplete information, integrity constraints, etc. The second problem deals with the need to represent and manipulate descriptive knowledge in the form of text, image, etc.; hence the need arises for support of multi-media information (see Akscyn et al., 1988).

The most efficient paradigm that could support this variety of information and facilitate the representation of multi-media in a consistent way seems to be the object-oriented one (see Kim, 1990). In order for a database management system to be object-oriented it should satisfy the basic features of a DBMS like: persistence, i.e. the ability of objects to persist in different program invocations; transactions, i.e. execution units that are executed either entirely or not at all; concurrency control, i.e. the algorithms that control the concurrent execution of transactions; recovery, i.e. the ability to recover from transaction, system, or media errors; querying mechanisms, i.e. the incorporation of high-level declarative constructs that let users quantify what they want to retrieve from the database; versioning, i.e. the ability to store and retrieve multiple versions of the same database object; integrity, i.e. the predicates that specify and define consistent states of persistent databases; security, i.e. the mechanism that control the user access rights; and performance issues, i.e. constructs and strategies for the enhancement of response time and throughput (see Kim, 1990). These features, however, should be implemented using the major object-oriented characteris- tics like: composite objects; user-definable types; object identity; encapsulation; types/classes; type/class hierarchies; overloading, overriding and late binding; and computational completeness (see Dittrich, 1990).

There are two major advantages of object-oriented databases: referential sharing and concurrent sharing. In referential sharing multiple applications and products share a common object; for instance, the same database table is part of a spreadsheet and a word processor. Object-oriented databases allow structuring and referential sharing of objects through supporting object identity and inheritance. On the other hand, like most conventional database systems, object-oriented databases control concurrent access to persistent objects by multiple users or applications.

Recent research in the area of intelligent and cooperative database systems suggest that the adoption of knowledge-based extensions of object-oriented database systems can solve the majority of problems in a consistent framework (see e.g. Manola et al., 1992; Papazoglou et al., 1992).

Model management

It has long been recognised that better computer-based systems are needed to support modelling in analytical organisations, where there are many models and many users. The typical model management system should support the following tasks: model formulation (i.e. gathering of information for building up a new model); model representation (i.e. representation of model structure, input and output information, overall characteristics, relations to other models, etc.); and model processing (i.e. application of suitable algorithms, support of sensitivity and what-if analyses, etc.), see Baldwin et al. (1991) for a survey in model management research.

Research in the model management context has seen the work of Geoffrion (1987, 1989a) with the structured modelling approach, which has been formally described in Geoffrion (1989b) and rigorously compared to well known modelling environments for operations research models, like GAMS (see Kendrick and Krishnan, 1989). This work extends operations research modelling to incorporate data modelling and symbolic reasoning. Another approach has been exactly the reverse, i.e. the use of the paradigm of symbolic reasoning and its extension and integration into management science, see e.g. Lee and Krishnan (1990) and Krishnan (1990) who use a predicate logic approach for a model management framework. A third line of thought has been the use of A I techniques for model management (see Dolk and Konsysnki, 1984; Jarke and Radermacher, 1988). Finally, the use of the relational and entity-relationship approaches has been explored for the design and management of models (see e.g. Muller-Merbach, 1983, 1990; Blanning, 1985, 1986).

In general, research in this area has been partial, in the sense that its aims were in supporting specific phases of the modelling life-cycle. Only recent efforts have been directed toward an integrated

International Transactions in Operational Research Vol. I , No. 2 179

support of mathematical modelling. The adoption of an object-oriented knowledge environment seems to be the most promising solution for the representation of multiple models as well as a unifying framework for the resolution of the following dichotomy that generally appears in the literature of model management: a model is either a procedure that accepts input and produces output (e.g. an LP algorithm is the model), or a model is the input to a procedure (i.e. the equations that describe input to the LP procedure constitute the model and the algorithm is the solver) (see e.g. Holsapple and Whinston, 1990).

A four-level hierarchy abstraction is often used in order to facilitate the study of model classification (see also Geoffrion, 1989a). This hierarchy would be the following:

different ‘specific models’ within a single ‘model class’; a specific model being a definite instance of a model class, e.g. a specific model could be a simulation that considers a serially routed machine-constrained job shop. different ‘model classes’ within a single ‘modelling paradigm’, e.g. a ‘model class’ could be a collection of all conceivable similar scheduling models. different ‘modelling paradigms’ within a ‘modelling tradition’; a ‘modelling paradigm’ is a collection of similar model classes, e.g. a modelling paradigm could be the collection of all job-shop scheduling models. different ‘modelling traditions’ within scientific fields; since ‘modelling traditions’ refer to the distinct modelling traditions our example would refer to modelling sequencing and scheduling problems in general.

As is obvious, such a taxonomy can make use of the object-oriented features of classification, generalisation and specialisation. Since the model is considered here as a quantitative abstract representation of reality that will be the input to an evaluation process, we should define an object-oriented structure for the taxonomy of evaluation procedures; the ‘model solvers’. Model solvers can be classified in a similar manner as models.

Furthermore, the object-oriented paradigm can provide the means for resolving the organisational and programming issues related to handling large model-bases as well as large solver-bases. Assume that as one of the attributes of each model instance, one defines the identification of a corresponding solver that is able to produce a numerical result; then handling and interfacing a model-base and a solver-base can be reduced to the problem of handling message-passing between various object classes.

On the other hand, knowledge representation mechanisms can be used both for the mathematical model formulation and representation tasks [see Krishnan et al. (1992) and Krishnan and Bhargava (1992) for a survey of the usage of artificial intelligence techniques in computer-aided model construction].

Knowledge management

Knowledge-based (management) systems and inference engines have recently gained more and more importance. Six types of knowledge have been identified, irrespective of the specific techniques for knowledge representation and processing; descriptive knowledge; procedural knowledge; presenta- tion knowledge; assimilative knowledge; linguistic knowledge; and reasoning knowledge (see also Holsapple and Winston, 1989).

In the case of reasoning knowledge the representation schemes used include: production rules (fragments of reasoning knowledge consisting of at least a premise and a conclusion); semantic networks (consisting of nodes, objects and directed, named edges); and frames (structures/schemata which contain properties/values and to which procedures can be attached). Inference engines in knowledge-based systems perform control tasks using: forward reasoning (forward chaining); reverse reasoning (backward chaining or goal directed reasoning); blackboards (shared main memory); pattern-direction; etc. (see Altenkrueger, 1990). Search techniques usually employed include: best-first; exhaustive search; breadth-first (FIFO); depth-first (LIFO); different modes of backtracking; etc., while reasoning techniques include: deductive and inductive inferencing; monotonic and non-monotonic reasoning; negations and default reasoning; etc.

180 Gregory Mentzas-Intelligent Organisational Information Systems

It has been argued that all knowledge-based systems require an external denotational semantics (see Patel-Schneider, 1990). If there is no such semantics for a system, then the system cannot be said to represent anything. Actually, many knowledge-based systems deal with facts as sets of unconnected sentences and are based on some assertional logic, such as first-order logic. On the other hand, the same rationale that motivates object-oriented programming can motivate an object-oriented approach in knowledge representation.

Hence, an object-oriented knowledge representation system would be a knowledge system that represents knowledge in terms of structured objects and classes. Such a system could extend the representational ideas found in the traditional artificial intelligence approaches, based on an external semantics that would provide meaning to objects, classes, roles and their related notions. Actually, the entire knowledge space can be represented as a node-and-link space where nodes represent either abstract concepts or specific objects. The links can represent inheritance relationships (is-a), attributes (has-a) and other complex or general relationships (kind-of, type-of, related-to, etc.) (see Patel-Schneider, 1990).

Cooperation support

Decision making is hardly the task of an individual. Conceptualised as the essence of modern organisations, groups constitute a key basis for decision making and acquisition of knowledge on organisations. The study of group meetings provides insight into group processes and the relationship between group cohesion and task performance (see DeSanctis and Gallupe, 1987). Recent advances in information technology, communications and management science techniques have enhanced the performance of systems that support multiparticipant processes (see Applegate et al., 1991). An important issue in multiparticipant processes deals with the assignment of quantitative values from group members as collective inputs to models which then help arriving at a solution; multi-attribute value analysis and calculation of preferences are techniques based on input collected in a group context. Methods used for deciding among alternatives include optimisation techniques, payoff matrices, utility curves, decision trees, game theory, ranking and statistical inference.

One may differentiate between two types of problems in the support of multiple interactive participants engaged in a joint production task: support of communication between team members; and structuring of the decision process (see Pinsonneault and Kraemer, 1989; Applegate, 1991). The main purpose of communication support is to reduce communication barriers in groups. The most important facilities used include information control (storage and retrieval of data), representational facilities (plotting and graph capabilities, large video displays) and group collaboration support facilities for idea generation, collection and compilation. Implementations that handle such issues include teleconferencing, electronic boardrooms and local group networks (see Kraemer and King, 1988). The structuring of decision making within a group is mainly implemented by the use of DSS for individuals, without support of the group per se.

A significant extension of traditional artificial intelligence systems has been the study of expert cooperation for distributed problem solving (see Huhns, 1987; Gasser and Huhns, 1989). In this approach, communication and negotiation is considered between multiple agents, which exhibit expertise in specific problem areas. An object-oriented approach (coupled with knowledge-based extensions) could facilitate both the organisation of experts in hierarchical structures, as well as guarantee their consistent communication, via e.g. the use of message-passing mechanisms.

Human interaction

Woods and Roth (1988) provide three design metaphors for man-machine systems: man as cooperating participant of a problem-solving/decision-making system; man as a supervisor of a technical system that is partly automatically controlled; man as a user of tools, this includes the use of communication tools which could be any type of technology-mediated human-human interaction. These metaphors can be regarded as orthogonal design dimensions. Research efforts in human- computer cooperative work would involve the cooperation dimension (with the development of

International Transactions in Operational Research Vol. I , No. 2 181

intelligent interfaces), the supervision dimension (i.e. in association with control applications involving human and system agents) and the tool dimension (for facilitating human-human cooperative work with intelligent tools) (see e.g. deGreef et al., 1991).

Research related to the cooperation dimension includes not only aspects of effective technological support, but also aspects of user feedback. In addition to keyboards and screens, interface devices include alternative methods that can support human-machine interaction and present information in a ‘user-friendly’ manner. Frameworks for modelling user-computer interaction have been developed, and they suggest there are possibilities for developing a language for information presentation and elicitation in the user-computer dialogue process. Opportunities have been identified to explore the areas of embedding facilitator expertise and aspects of system usage experience in the knowledge base and of creating on-line process monitoring (see Nunamaker et al., 1989).

The recent trend towards separating the dialogue and computation components of software raises a lot of problems. Some of them are concerned with the appropriate definition of dialogue (as what the user perceives), the functional semantics of applications, the role of the user interface management system in the overall system architecture, etc. However, the major issues are focused on the choice over possible interaction styles and their requirements for control and communication (see Hartson, 1989).

There are two main interaction styles: the conversational style and the model style. In the conversational approach the user describes what to do with a command language, while in the model approach the user manipulates visual representations of things. In the model approach direct manipulation is often associated with multi-thread dialogue, which means that a user has many task paths available at any time during the dialogue. In order to support such a dialogue, the underlying system cannot be successfully managed by ordinary sequential (synchronous) control; it requires asynchronous control mechanisms. While sequential dialogue control has a very strong linguistic aspect and can be associated with control structures based on some variation of deterministic finite-state machine, asynchronous dialogue models are not connected linguistically. Each token is connected with an event that results from a user action, causing some system action in response. An example of an event is the sensing of the mouse cursor passing over an icon. Actually, event-based mechanisms have the greatest descriptive power in user interface management systems (see Green, 1986).

An object-oriented approach to dialogue support provides significant advantages, since it has inherent features that exploit event-driven communication (via message-passing) and encapsulation. A window, for example, would be an instance of the window object class, while window management can be easily accomplished by sending messages to a window and it will perform the necessary operations on itself. Event handlers can be modelled as object classes, while dialogue concurrency can be supported with more general cases of event-based processing, like the one found in multi-agent systems, see Agha (1990) and Agha and Hewitt (1987), for a theoretical discussion.

INTELLIGENT ORGANISATIONAL INFORMATION SYSTEMS

We believe that effective enterprise-wide decision making could be greatly facilitated by the existence of software entities with autonomous processing capabilities, which own a private data- and knowledge-base, and which act on their environment on the basis of information they receive, perceive, process, retain and recall. We label such elements active intelligent objects (or more simply ‘objects’), and claim that such objects could provide the basic building blocks for intelligent organisational information systems.

In this section we give an abstract and behavioural definition of active intelligent objects, since different authors have interpreted similar concepts in different ways, while all of them have tried to capture a notion of object intelligence, and of the object as an encapsulation of an asynchronous locus of activity. Finally, we present a conceptual message-based architecture, that could provide the initial architecture for designing 101s.

In an attempt to avoid giving a formal definition of active intelligent objects, we prefer to list some of their most crucial features. We classify these features in two groups that correspond to the two

182 Gregory Mentzas-Intelligent Organisational Information Systems

different levels of the system designed: the level of the single, independent element (i.e. the internal structure of the object), and the system level (i.e. the system architecture of TOIS). The features referring to the internal structure are: specialisation; representation; effectuation; learning; adaptability; planning; and intentionality. The features of the system architecture are: parallelism; distribution; modularity; heterogeneity; communication; organisation; and human interaction [see Mentzas (1993b) for a detailed presentation]. Table 5 gives an overview and an explanation of these features. An illustration of the operation of 101s systems in a production environment is also given in Mentzas (1993b).

Table 5 . Requirements of computational systems for intelligent organisational decision making

Requirements Explanation

Internal structure Specialisation

Representation

Effectuation or transformation

Learning

Adaptation

Planning

Intentionality

System architecture Parallelism

Distribution

Heterogeneity

Modularity

Communication

Organisation

Human interaction

Ensures that each computational element can efficiently solve a part of a given problem, i.e. that it has a specific area of expertise. Attempts to capture the requirement that each element represents semantically loaded information. For example, transformation of a document in an office, raw material in a factory, etc. Difference between reactive and deliberative effectuation. Ability of computational elements to perform new tasks that could not be performed before. By adapting, a system changes its current behaviour, in order to account for newly acquired information and perform old tasks better. Planning capabilities are needed, when prior to taking action, a data structure representing the intention to take the action is developed. Intentionality is defined as that feature by which mental states are directed at or about other computational elements and states of affairs in the world.

Philosophy of breaking a problem to be solved into manageable tasks and allocating these tasks across several computational elements. Corporations of a significant size face the problem of ensuring consistency and correctness of information residing at various physical sites or processors. Need for the consistent integration of heterogeneous behaviour, i.e. computational elements with different specialisations and structures. Design of a system using concrete basic blocks, whose interconnection supports a modular approach. The glue that makes cooperation possible; use of common access methods that can be implemented independently from internal representations. The modelling of authority structure, groupings, coalitions of computational elements, etc. The modelling of human actors in a similar manner as system objects and the interactions between them.

An active intelligent object can be generally defined as an object in which a high degree of autonomous responsibility and control is vested. Independent objects can be considered as agents, which are sources of knowledge and activity (see also Ellis and Gibbs, 1989; Grant, 1991). Each individual object is characterised by specific domains of expertise, has the capacity to solve complex problems and can work independently for problems tailored to its contextual subject matter. Information sharing and information exchange are required to allow systems of objects to create consistent views of problems and arrive at right solutions. 101s systems are groupings of active intelligent objects which coordinate their knowledge, goals, skills and plans jointly to take action or to solve problems, i.e. in an organisational setting they behave like multi-agent systems (Bond and Gasser, 1988).

Although the complexity of each active object can be reduced, the overall complexity of the system must be considered. A planning capability should be introduced, so that a global problem addressed to the 101s system could be decomposed into a variety of sub-problems and subsequently each sub-problem should be distributed to the relevant independent object. In addition to such ‘query (or task) decomposition’ facilities, there is also the need for ‘answer synthesis’ strategies that would reconcile partial solutions and rationalise the information flows and the answers to a particular problem.

Such planning facilities should be opportunistic and take cognisance of things such as those activities which are taking place, and which active object is performing these activities. There has

International Transactions in Operational Research Vol. I , No. 2 183

been some discussion in the distributed artificial intelligence literature over the benefits and drawbacks of global and local planning facilities. The former seems to be more appropriate for less dynamic environments than decision making in organisations, while the latter, although they allow optimal usage of agents, do not guarantee overall optimal usage (see Davis and Smith, 1983). Hence a two-level planning capability seems to be the most appropriate (see also O’Hare, 1990).

In order to alleviate the problem of cooperation between agents, hierarchical structures may be employed; inter-object control flows downwards in this structure and information flows upwards. Local partial solutions are interfaced and combined with those of other active objects solving dependent tasks; such a system resembles the ICIS agents proposed in Papazoglou et al. (1992). Such a structure, however, raises the need for consistent cooperation in a decentralised environment. Decentralisation implies that both control, knowledge and data are logically and spatially distributed. The system lacks global control, as well as global data storage; hence, no independent object has either a global view of the problem examined, or a global view of the activities carried out within the overall system. In addition, 101s systems are loosely coupled, in the sense that each object is mainly occupied in processing individual computational activities and not in communicating with other objects, except when this is necessary.

Structure of I O I S objects

Each constituent element of an 101s (active intelligent object) should consist of a local knowledge base, a deductive capability, a planning facility and a communication mechanism, that would enable it to interact with other objects of its community. Consequently each knowledge base would be relatively small, and the inference engine which operates upon this corpus of knowledge would not require a high degree of sophistication. Independent objects are modelled as active software elements capable of reasoning to external stimuli. The latter may be requests for information, processing, reasoning or decision making.

The knowledge base of each intelligent object models two types of knowledge: the local area of expertise, in which the specific object and its close acquaintances are ‘experts’; and the types of expertise in which remote objects are ‘experts’. The knowledge base includes detailed information about the former, but only partial and abstract knowledge about the latter.

Communication with other objects is enabled with the exchange of messages. These may be of two general types: specific requests to proceed with the solving of a problem, in which the current object is an ‘expert’, or requests to solve problems in which the current object is not an ‘expert’. In the latter case, a negotiation mechanism is initiated; this is based on the contract net protocol; see the following subsection.

Cooperation and negotiation in I O I S

In order to explicitly provide facilities for communication and negotiation in an organisational setting, we need to study cooperation facilities.

Active intelligent objects are organised in clusters, in terms of their area of expertise. One useful organisation is to group objects in corporate divisions, similar to the divisions of actual enterprises. Hence, one would be expected to form communities of marketing objects, strategic planning objects, scheduling objects, etc. It is interesting to note that such an architecture would (relatively) easily facilitate the incorporation of already available knowledge and inferencing mechanisms for specific enterprise domains. Hence, the wealth of existing expert system applications for relevant parts of an enterprise could be re-used, after some modifications, that would permit their incorporation in the object system. Although functional decomposition can be the organisational principle for structuring intelligent organisational information systems, there is no need to assume the existence of a one to one mapping between functions and specific objects. It would seem more natural to expect that such a mapping is one to N (see also O’Hare, 1990).

A hierarchical structure is adopted for the organisation of object groups into societies, similar to the one proposed by Huhns et al. (1983). Such a hierarchical tree-like structure can be achieved by

184 Gregory Mentzas-Intelligent Organisational Information Systems

successive functional decompositions; in this manner, groups of objects are permitted to communi- cate only with their immediate ancestor, descendants and objects at the same level of the particular subtree.

Inter-object communication and task decomposition may take place by means of the well known in the distributed artificial intelligence research literature contract net protocol (CNP) (see Davis and Smith, 1983; Smith, 1980). CNP follows a negotiation scheme in which workers agents submit their bids on subtasks, or which they are suited, to a manager agent. The manager agent awards the contract for solving the subtasks to the most appropriate workers agent based on their bids. Contracting involves an exchange of information between agents, an evaluation of the information by each member from its own perspective, and a final agreement by mutual selection.

The negotiation protocol adopted for societies of objects (i.e. 101s systems) is similar with CNP. However, the hierarchical structure employed for the organisation of 101s generates two differences with CNP. First, the individual objects are bound by decisions of their superordinate objects in a negotiation, while agents in CNP are always free to exit from a contracting process. Second, global optimisation of subtasks is achieved by the manager agent; for a similar approach see also inter-agent cooperation of ICIS in Papazoglou et al. (1992).

CONCLUSIONS

This paper has argued that although specific types of computer-based information systems (CBIS) are powerful tools in certain parts of the decision and information-related processes in modern organisations, none of them provides integrated support for all processes. This can be attributed to their lack of advanced abilities such as perception, interpretation, reasoning, explanation, goal- setting, learning, etc., as well as the lack of multi-participant, organisational features as cooperation, conflict, negotiation, etc. We argue that a consistent synthesis of current research in such fields as computer science, operations research and artificial intelligence provides a potential solution to the satisfaction of these needs in a uniform way.

The main thesis of this paper is that the organisational decision-making functions can be facilitated by the use of intelligent software entities with autonomous processing capabilities, enhanced with coordination and negotiation facilities. We have labelled systems that comprise such entities intelligent organisational information systems (101s). The background technology for 101s is based on object-oriented computing and distributed artificial intelligence. Distributed AI platforms, such as MACE (Gasser et al., 1987) and knowledge-based extensions of object-oriented technologies, as in Manola et al. (1992), may prove to be the starting vehicles for 101s implementation.

Such an approach would generate increased benefits. First, 101s cooperation can provide an interesting metaphor for the natural interaction among human experts in various fields. Second, the conceptual architecture presented here supports incremental design, modularisation and functional decomposition; thus, it can assist in step-wise implementation, testing and refinement. Third, the organisational aspects of the architecture (i.e. loosely coupled decentralisation and message-based communication) can offer increased reliability, since they ensure degradation of performance in the case of intelligent objects that fail to achieve their goal. Finally, the abundance of existing knowledge-based applications could be reused, after some appropriate modifications.

Research towards intelligent organisational information support nevertheless raises many unsolved issues. First, we need to support vast networks of heterogeneous, distributed computa- tional elements; this need calls for the introduction of broadband telecommunication and computing networks with new requirements for intelligent systems of network control. Second, parallel processing facilities should be supported; this issue is directly connected to the use of multiprocessor computer architectures. Optimisation issues and run-time object creation issues arise in this context. Finally, the areas of distributed artificial intelligence, distributed databases and concurrent object-oriented programming should be further explored, in order to provide consistent frameworks for active objects.

Acknowledgement - The author would like to thank Professor Rolfe Tomlinson and the reviewers for helpful comments on an earlier draft.

International Transactions in Operational Research Vol. I , No. 2 185

REFERENCES

Ackoff, R. L. (1967). Management misinformation systems. Management Science, Vol. 11, No. 4, pp. 147-156. Agha, G. A. (1986). Actors: A Model of Concurrent Computation in Distributed Systems. Cambridge, MA: MIT Press. Agha, G.A. (1990). Concurrent object-oriented programming. Communications of the A C M , Vol. 33, No. 9, pp. 125-141. Agha, G. A. & Hewitt, C. (1987). Actors: a conceptual foundation for concurrent object-oriented programming. In B. Shriver

Akscyn, R. M., McCracken, D. L. and Yoder, E. A. (1988). KMS: a distributed hypermedia system for managing knowledge

Altenkrueger, D. E. (1990). KBMS: aspects, theory and implementation. Information Systems, Vol. 15, No. 1, pp. 1-7. Alter, S. (1980). Decision Support Systems: Current Practice and Continuing Challenges. Reading, MA: Addison-Wesley. Applegate, L. (1991). Technology support for cooperative work: a framework for studying introduction and assimilation in

Applegate, L., Ellis, C., Holsapple, C., Radermacher, F. & Whinston, A. (1991). Organizational computing: definitions and

Baldwin, A. A., Baldwin, D. & Sen, T. (1991). The evolution and problems of model management research. Omega, Vol. 19,

Barbic, F., Ceri, S . & Braachi, G . (1985). Modeling and integrating procedures in office information systems design.

Blanning, R. (1985). A relational framework for joint implementation in model management systems. Decision Support

Blanning, R. (1986). An entity-relationship approach to model management. Decision Support Systems, Vol. 2, No. 1, pp.

Bonczek, R. H., Holsapple, C. & Whinston, A. B. (1981). Foundations of Decision Support Systems. New York: Academic

Bond, A. H. & Gasser, L. (Eds) (1988). Readings in Distributed ArtiJcial Intelligence. San Mateo, CA: Morgan Kaufman. Ching, C., Holsapple, C. & Whinston, A. B. (1993). Modeling network organizations: a basis for exploring computer support

Davis, R. & Smith, R. (1983). Negotiation as a metaphor for distributed problem solving. ArtiJiciaf Intelligence, Vol. 20, No. 1,

deGreef, P., Mahling, D., Neerincx, M. & Wyatt, S. (1991). Analysis of human-omputer cooperative work. Technical Report

DeSanctis, G. & Gallupe, R. B. (1987). A foundation for the study of group decision support systems. Management Science,

Ditrich, K. (1990). Object-oriented database systems: the next miles of the marathon. Information Systems, Vol. 15, No. 1, pp.

Dolk, D. R. & Konsysnki, B. (1984). Knowledge representation for model management systems. IEEE Transactions on

Durfee, E. H. (1988). Coordination of Distributed Problem Solvers. Boston: Kluwer. Edwards, J. S. (1992). Expert systems in management and administration - are they really different from decision support

systems? European Journal of Operational Research, Vol. 61, pp. 1 14-121. Ellis, C.A. & Gibbs, S . J. (1989). Active objects: realities and possibilities. In W. Kim & F. H. Lochovsky (Eds).

Object-Oriented Concepts, Databases and Applications. New York: ACM Press. Ellis, C. A. & Nutt, G. J. (1980). Office information systems and computer science. ACM Computing Surveys, Vol. 12, No. l ,

Ellis, C. A., Gibbs, S. J. & Rein, G. L. (1991). Groupware: some issues and experiences. Communications of the A C M , Vol. 34,

Eom, H. & Lee, S. (1990). Decision support systems applications research: a bibliography (1971-1988). European Journal of

Gasser, L. & Huhns, M. N. (1989). Distributed ArtiJiciaf Intelligence, Vol. 11. Los Altos, CA: Morgan Kaufmann. Gasser, L., Braganza, C. & Herman, N. (1987). MACE: a flexible testbed for distributed AI research. In M. N. Huhns (Ed.).

Geoffrion, A. M. (1987). An introduction to structured modeling. Management Science, Vol. 33, No. 5, pp. 547-588. Geoffrion, A. M. (1989a). Integrated modeling systems. Computer Science in Economics and Management, Vol. 2, No. 1, pp.

Geoffrion, A. M. (1989b). The formal aspects of structured modeling. Operations Research, Vol. 34, No. 1, pp. 30-51. Grant, T. J. (1991). A review of multi-agent systems techniques, with application Columbus user support organization. Future

Green, M. (1986). A survey of three dialog models. ACM Transactions on Graphics, July, pp. 244-275. Hartson, R. (1989). User-interface management control and communication. IEEE Software, Vol. 6, No. 1, pp. 62-70. Hatcher, M. E. (1992). Group decision support systems: decision process, time and space. Decision Support Systems, Vol. 8,

Hewitt, C. (1977). Viewing control structures as patterns of passing messages. Artijicial Intelligence, Vol. 8, pp. 323-364. Holsapple, C. & Winston, A. B. (1989). Knowledge representation and processing in economics and management. Computer

Holsapple, C. & Winston, A. B. (1990). Toward an environment theory of decision support. Computer Science in Economics

Huber, G. P. (1982). Organizational information systems: determinants of their performance and behavior. Management

Huhns, M.N. (1987). Distributed Artijicial Intelligence. Los Altos, CA: Morgan Kaufmann. Huhns, M. N., Stephans, L. M. & Bonnell, R. D. (1983). Control and co-operation in distributed expert systems. Proceedings

& P. Wegner (Eds). Research Directions in Object-Oriented Programming (pp. 49-74). Cambridge, MA: MIT Press.

in organizations. Communications of the A C M , Vol. 31, No. 7, pp. 820-835.

organizations. Journal ofOrganizationa1 Computing, Vol. l, No. 1, pp. 11-39.

issues. Journal ofOrganizationa1 Computing, Vol. 1, No. 1, pp. 1-10.

NO. 6, pp. 51 1-528.

Informarion Systems, Vol. 10, No. 2, pp. 149-168.

Systems, Vol. 1, No. 1, pp. 69-82.

67-72.

Press.

coordination possiblities. Journal of Organization Computing, Vol. 3 (forthcoming).

pp. 63-109.

No. 4, Esprit Project 5362, Leiden University, Leiden.

Vol. 33, pp. 589-606.

161-167.

Software Engineering, Vol. SE-10, pp. 609-628.

pp. 27-60.

NO. 1, pp. 39-58.

Operational Research, Vol. 46, pp. 333-342.

Distributed ArtiJcial Intelligence (pp. 119-152). London: Pitman.

3-15.

Generation Computer Systems, Vol. 7, pp. 413-437.

NO. 2, pp. 83-84.

Science in Economics and Management, Vol. 2, No. 1, pp. 37-48.

and Management, Vol. 3, No. 1, pp. 81-105.

Science, Vol. 28, pp. 138-155.

of the IEEE Southeast Conference.

186 Gregory Mentzas-Intelligent Organisational Information Systems

Hynynen, J. (1992). Using artificial intelligence technologies in production management. Computers in Industry, Vol. 19, No.

Jarke, M. & Radermacher, F. (1988). The AI potential of model management and its central role in decision support. Decision

Keen, P. G . W. (1987). Decision support systems: the next decades. Decision Support Systems, Vol. 3, pp. 253-265. Keen, P. W. G . & Scott-Morton, M. S. (1978). Decision Support Systems: An Organisational Perspectioe. Reading, MA:

Kendrick, D. & Krishnan, R. (1989). A comparison of structured modeling and GAMS. Computer Science in Economics and

Kim, W. (1990). Object oriented databases: definitions and research directions. IEEE Transactions on Knowledge and Data

Kraemer, K. L. & King, J. L. (1988). Computer-based systems for cooperative work and group decision making. ACM

Krishnan, R. (1990). A logic modeling language for automated model construction. Decision Support Systems, Vol. 6, pp.

Krishnan, R. & Bhargava, H. (1992). Computer-aided model construction. Decision Support Systems, Vol. 9, pp. 91-1 11. Krishnan, R., Li, X. & Steier, D. (1992). Development of a knowledge-based mathematical model formulation system.

Lee, R. & Krishnan, R. (1990). Logic as an integrated modeling framework. Computer Science in Economics and Management,

Li, Y. P. (1985). On data modelling through logic. Research Paper, Graduate School of Management, UCLA, 8 February. Manola, F., Heidler, S., Georgakopoulos, D., Hornick, M. & Brodie, M. (1992). Distributed object management.

Markus, M. L. & Robey, D. (1988). Information technology and organizational change: causal structure in theory and

Matsuda, T. (1993). Utility of OR/MS models in converting information into intelligence. Paper presented in the IFORS '93,

Mentzas, G . (1991). A review of object orientation and knowledge processing in office models. European Journal of

Mentzas, G. N. (1993a). Coordination of joint tasks in organizational processes. Journal of Information Technology, Vol. 8,

Mentzas, G. (1993b). Object-based intelligence in office and production processes: a view on integration. Information and

Muller-Merbach, H. (1983). Model design based on the systems approach. Journal of the Operational Research Society, Vol.

Muller-Merbach, H. (1990). Database-oriented design of planning models. I M A Journal ofMathematics Applied in Business & Industry, Vol. 2, pp. 141-155.

Nunamaker, J., Vogel, D. & Konsynski, B. (1989). Interaction of task and technology to support large groups. Decision Support Systems, Vol. 5, No. 2, pp. 139-152.

O'Hare, G. M. P. (1990). Designing intelligent manufacturing systems: a distributed artificial intelligence approach. Computers in Industry, Vol. 15, pp. 17-25.

Papazoglou, M. P., Laufmann, S.C. & Selis, T. K. (1992). An organizational framework for cooperating intelligent information systems. International Journal of Intelligent and Cooperative Znforrnation Systems, Vol. 1 , No. 1 , pp. 169-202.

Parker, B. & Al-Utaibi, G. A. (1986). Decision support systems: the reality that seems hard to accept? Omega, Vol. 14, No. 2,

Partridge, D. (1987). The scope and limitations of first generation expert systems. Future Generation Computer Systems, Vol. 3,

Patel-Schneider, P. (1990). Practical, object-based knowledge representation for knowledge-based systems. Information

Pinsonneault, A. & Kraemer, K. (1989). The impact of technological support on groups: an assessment of the empirical

Quarterman, J. & Carl-Mitchell, S. (1993). The computing paradigm shift, Journal of Organizational Computing, Vol. 3

Reiter, R. (1985). Towards a logical reconstruction of relational database theory. In M. L. Brodie, J. Mylopoulos & J. Schmidt

Rockart, J. F. & DeLong, D. W. (1988). Executive Support Systems. Homewood: Irwin. Rose, T., Maltzahn, C. & Jarke, M. (1992). Integrating object and agent worlds. Presented at the 4th Conference on Advanced

Shaw, M. E (1981). Group Dynamics: The Psychology of Small Group Behauior, 3rd edn. New York: McGraw-Hill. Simon, H.A. (1977). The New Science of Management Decision, rev. edn. Englewood Cliffs, NJ: Prentice-Hall. Smith, R. (1980). The contract-net protocol: high level communication and control in a distributed problem solver. I E E E

Sprague, R. H. & Carlson, E. D. (1982). Building Eflectioe Decision Support Systems. Englewood Cliffs, NJ: Prentice-Hall. Sprague, R. H. (1980). A framework for the development of decision support systems. M I S Quarterly, Vol. 4, pp. 1-26. Turban, E. (1990). Decision Support and Expert Systems: Management Support Systems. New York: Macmillan. Turban, E. & Trippi, R. (1989). The utilization of expert systems in OR/MS: an assessment. Omega, Vol. 17, No. 14, pp.

1 , pp. 21-35.

Support Systems, Vol. 4, No. 4, December.

Addison-Wesley.

Management, Vol. 2, No. 1 , pp. 17-36.

Engineering, pp. 327-340.

Computing Surveys, Vol. 20, No. 2, pp. 115-146.

123-1 52.

Communications of the A C M , Vol. 32, No. 9, pp. 138-146.

Vol. 3, NO. 1 , pp. 107-136.

International Journal of Intelligent and Cooperative Information Systems, Vol. 1 , No. 1 , pp. 5 4 2 .

research. Management Science, Vol. 34, No. 5 , pp. 583-598.

XI11 World Conference on Operations Research, 12-16 July. Lisbon, Portugal.

Information Systems, Vol. 1 , No. 3, pp. 193-203.

pp. 139-150.

Decision Technologies, forthcoming issue.

34, NO. 8, pp. 739-751.

pp. 135-143.

pp. 1-10.

Systems, Vol. 15, No. 1, pp. 9-19.

research. Decision Support Systems, Vol. 5 , No. 2, pp. 197-216.

(forthcoming).

(Eds). On Conceptual Modelling. Berlin: Springer.

Information Systems Engineering, CAISE '92, 12-15 May, Manchester.

Transactions on Computers, Vol. C-29, No. 12, pp. 1104-1 113.

31 1-322. Turban, E. & Watkins, P. (1986). Integrating expert systems and decision support systems. M I S Quarterly, Vol. 10, No. 2, pp.

120-1 36. Vogel, D. & Nunamaker, J. (1988). Group decision support system impact: multimethodological exploration. In J. Galegher,

R. Kraut & C. Egido (Eds). Proceedings of Conference on Technology and Cooperatioe Work (Tucson, Arizona, 25-28 February). National Science Foundation, Bell Communications, University of Arizona, Tucson.

Watson, H., Rainer, R. K. & Koh, C. (1991). Executive information systems: a framework for development and a survey of

International Transactions in Operational Research Vol. I , No. 2 187

current practices. M I S Quarterly, Vol. 15, No. 1, March. Watson, H. J &L Hill, M. M. (1983). Decision support systems or what didn’t happen with MIS. Interfaces, Vol. 13, No. 5 , pp.

8 1-88. Williams, C. (1986). Expert systems, knowledge engineering and AI tools - an overview. IEEE Expert, Winter, pp. 66-70. Woods, D. D. & Roth, E. M. (1988). Cognitive engineering: human problem solving with tools. Human Factors, Vol. 30, pp.

Zennetos, Z. S. & Sertel, M. R. (1970). Computerized management information systems and organization structures. Working 41 5-430.

Paper 486-70, Sloan School of Management, MIT.