Peace Games with Open Modeling Network

ABSTRACT

In the last few years the transformation of the communication network, which started with launching of the first communication satellite in the mid 1960s, culminated in establishment of globally interconnected packet switched Value Added Networks (VANs). Concurrently with this technological development we can observe growing and unsolved difficulties in dealing with the problems caused by the population explosion, depletion of natural resources and problems of global ecology management.

On the other hand, recent trend to automation, which is fueled by economic competition in particular between the U.S. and Japan, development of expert systems and the explosive growth of defense systems with shorter and shorter reaction time are rapidly changing the global economic and political situation in both developed and developing countries. The growing interdependence of national economies and the complexity of the global issues require higher level of cooperation and understanding between the highly diverse groups and nations which populate this planet.

In this article we will examine the application of the new development in the area of distributed systems and Computer Aided Communication (CAC) to the analysis of the global sociological and economical issues. Based on the review of the past attempts and experiences with model acceptance and validation, we argue that meaningful and credible simulation has to be implemented as a modeling network composed of a large number of locally developed and verified models. No single model, developed by local group of experts has a chance for universal acceptance when it is dealing with controversial and confrontation prone area such as global resource allocation and economical policies.

Yet, a comprehensive model of global resources, ecology and economy is needed for the rational management of ecology and for economic cooperation between nations and economic blocks. As a solution to the dilemma between the need for a unified model and a diversity of views and the special interests of diverse groups, we consider a public Open Modeling Network (OMN) which will consist of models developed by local experts interconnected by global VANs.

New problems of interfacing of models which utilize different methodologies and use different computers and computer languages will require adoption of new standards, allowing for translation of the content and meaning of the information generated and needed by the individual models and development of policies which would prevent manipulation of the data by special interest groups. In the long run, it can be expected that the benefits of participation in the public network will exceed the problems caused by sharing some data and costs. The benefits will result from access to vast databases of relevant and up-to-date economical information and improved communication on the global scale.

The problem of managing the variety of heterogeneous models, each operating locally, yet affected from time to time by the results of similar runs at other locations, is compared to Scheduling Algorithm problem, which is required by all asynchronous distributed systems consisting of the distributed communicating processors. In particular we consider the application of Time Warp algorithm.

The GLObal Systems Analysis and Simulation (GLOSAS) Project proposes to utilize the semantic benefits of gaming simulation on a global scale to aid decision makers in appreciating the impact of their decisions on interwoven global problems, i.e., the construction of Globally Distributed Decision Support System (GDDSS) for plus sum, peace game. The CAC, with cooperative executions of autonomously managed simulation sub-models at distributed locations in gaming mode, can provide a "meta-language" allowing improved communications among users of the sub-models.

Progress in the study of distributed systems has yielded a new scheduling algorithm, the Virtual Time concept, which allows organization of the information exchange among dispersed, dissimilar computational resources with asynchronous and parallel executions. These new developments are applied here to the Distributed Computer Simulation Systems (DCSS) of the GLOSAS Project, which deals with coordination of the distributed sub-models and their experts via the global VANs for global crisis and ecology management.

1. INTRODUCTION: POWERFUL NEW TOOLS FOR COLLECTIVE INTELLIGENCE

Society needs much more sophisticated tools to deal with complex global problems, which so overwhelm the world's leaders that they are tempted to simplistic solutions. Nightingale (1985), writing about a playwright, spoke of Michael Frayn's concern for "the awesome complexity of the world, and our desperate attempts to reduce it to nice, neat shape." Gleick (1985) reported how the mathematician, Benoit Mandelbrot, has expanded the work of scholars who "missed a whole range of things" because they "simply didn't have the tools" they needed to deal with "complexity (which) has been developing slowly in many disciplines for nearly a generation." Mandelbrot's work, he said, is a part of the revolution in understanding chaos, the study of turbulence and disorder in a whole range of phenomena.

Now, however, powerful new computer-communication and simulation tools can make it possible, as never before in history, for any intelligent citizen to have a hand in developing new alternatives to war and other complex international problems. Even the political geniuses, and perhaps there are a few, have not been able to keep in mind all they need to know and understand to deal with the whole complexity of global inter-relation. But computers, combined with other electronic technology, can now make possible mind-tools for a powerful new "collective intelligence."

It is now possible to combine existing technologies to make sophisticated and more holistic explorations of various scenarios for solving global social problems. Many small computers in different countries can be interconnected, through globally distributed network processing and information processing, into modeling and simulation instruments as powerful as those used by the Pentagon for war gaming. People-enhancing and mind- empowering tools can thus be created by combining such technology as: Value Added Networks (VANs), satellites, packet radio, video disk and expert systems, global data banks, wireless portable terminals, and more.

Developing experiences from modeling and gaming can also be combined in global systems and data banks with a cascading effect, to empower explorations for new international institutions, remodeling existing ones, new strategies by official government agencies, universities, peace institutions, churches and lay person groups. Before they are put into effect, often at great cost and risk, new strategies can be explored through gaming simulations by collective effort of people with different views located at various parts of world, to see how they might prevent crises, deal with crises, and make various efforts more effective in preventing war and creating conditions for peace.

2. COMPUTERS PLUS WHAT?

Rossman (1985) describes various advanced tools that might be interconnected for powerful explorations through and for collective intelligence. They are:

(1) The meshing of phone and computer systems into a single mode, combined with expert systems and data banks via satellites creates a new tool with breathtaking possibilities. Computer expert systems, as intelligent assistants, can fuse the knowledge of many specialists into tools to deal with complex problems, as providing their users with diagnosis and hinted solutions, which often outsmart human experts who designed and built the systems.

(2) The work of one huge computer can be done by a distributed network of many interconnecting microcomputers, which make up a reasoning system, stocked with all necessary knowledge. Access to information stored on optical video disks, with high-powered laser diode, can be obtained within seconds, e.g., over one dozen volumes of encyclopedia can be packed into a single shiny 5 1/4 inch disk - even including color illustration diagrams and pictures, and very possibly with voice and music annotations in the future.

(3) A global computer network can be a major new tool for coordinating resources - including brainware of project participants. Computer modeling and simulations to explore risks and possibilities then become a powerful tool for calculating the consequences of experimental change, by the people of different views and disciplines in various countries who created those cooperative simulation models.

(4) Fifth generation computer tools, instead of solving problems step-by-step, can break complex projects up into thousands of units, each to work simultaneously by different computers all over the world, the so-called distributed, asynchronous parallel processing as resembling numerous neurons in our human brain. The fusing of expertise through networks of minds can result as thousands of interconnected computers help people work simultaneously on different aspects of the same problem or project, particularly on the utmost crisis of human kinds, i.e., to prevent nuclear war and holocaust. "No matter where a nuclear conflict would begin on our planet and no matter who would initiate the first strike, whether or not a retaliatory strike would follow, the entire human race would share a common fate: no one can hope to survive a nuclear catastrophe," as said by Moiseev (1984), Director of Research at Computer Center of U.S.S.R. Academy of Science.

3. MOTIVATION AND FIELD OF APPLICATION

The topic of this paper is a technological fix for intractable global problems, such as hunger, poverty or arms race, involving many diversified groups covering a wide social and political spectrum. These different groups have different motivations and views of the issues and use different techniques, terminologies and concepts. In such situations, communication problems often prevent a smooth progress to solution of the issues. New techniques of the distributed computer simulation can provide planning and management methods which can overcome wide communication gaps.

3.1 Minus vs. Plus Sum Games

Not all problems can be solved by improved communications alone. In some intractable problems, parties in the issues take a position, which they are able to maintain and defend within the existing socioeconomic power structures. Such problems resemble the situation of two armies, which dig in and wait for the outbreak of hostilities. Such situations can usually be represented mathematically as a game with negative sum, in which non-cooperative strategies are rewarded and tend to be chosen as an optimal strategy by at least some major players. Such situations are not amenable to our technological solution and will not be discussed here.

In this paper we address a class of the social situations which can be described by a plus sum game, a game in which the optimal strategies require cooperation and exchange of messages rather than exchange of artillery fire. These situations are often intractable because of the large number of players having different views of the problem, different special interest and conflicting approaches or belief systems.

The term "player" is used here in the abstract sense of the game theory and may include agencies, organizations, corporations or industries as well as individuals. The players are characterized by their needs, goals and means of manipulating their environment. The players may be geographically scattered, use different (natural as well as computer programming) languages and have a wide range of physical and technological tools.

3.2 Computer Simulation of Social and Managerial Problems

Determination of an optimal strategy for complex social problems would be difficult, even if the issue were well described and agreed upon by the parties of the game. Moreover, in social problems the concepts are usually fuzzy, ill defined and often changing their meaning during the game. We will discuss tools, which enable the players to reach consensus and interactively find, maintain and coordinate the optimal strategy.

The new technological developments, which we will describe below, do not address any of the social problems directly. The problems can only be solved by the consensus or battle of the players. The techniques address the cause of the intractability, namely the management problems related to the languages and communications.

"Management" is the activity enabling human groups to achieve their collective goals. Many techniques have been developed and refined ranging from art to scientific discipline (Boettinger, 1975). Management science, as it implies, is the scientific approach to the analysis of management of enterprises or systems and their behavior whether for individual, communal, national or for global.

In contrast to physical and nonliving systems, social and living systems are hardly ever amenable to testing and "management science" has an inherent contradiction between the nature of "management art" and its "scientific" approach. The remedy to this is to create models, which simulate the system behavior, much as it is done by an airplane model in a wind tunnel. Computer simulation, which utilizes software rather than a hardware model, is therefore the essential tool of management science.

3.3 Importance of Modeling

Schank (1984) of the Yale Artificial Intelligence Lab points out that from now on it will be essential to use computer modeling for making important decisions, models which incorporate more and more knowledge about people and institutions. Until recently, he says, it has not been possible to make large conceptual computer models of governments, of the work of politicians and other complex systems. Now, however, such models can be increasingly complex, integrated, and can be more and more useful and trustworthy for testing ideas, theories and possible actions. Computers will not make good decisions but can be used to help human beings make better ones.

Licklider (1983) says that computer modeling and simulations are already beginning to play an important role in government research and planning, as these expand and multiply beyond space and military projects to other national planning efforts. The Soviet Union, he reports, is planning to create a 3,000-computer nationwide network with databases for planning. (Russians were, after all, the first who attempted to apply linear programming optimization to their national economic planning, albeit premature at that time.) Moiseev (1984) proposed that "further advances in the instruments of global analysis ... should be widely applied in arriving at quantitative characteristics of global processes and in evaluating the capacity of alternative development strategies to influence the course of human civilization."

Gilpin (1983), in discussing war games, says that the economic and military changes which result from the use of computers and other advanced technologies are bringing human society into an age wherein more is to be gained through cooperation and an international division of labor than through strife and conflict. For in the electronic global village all people will either lose or win together. To survive in a global society, Shubik (1983) suggests, we must develop tools to control pollution, fight inflation, provide justice and welfare, and to warn of new dangers and threats, such as acid rain or greenhouse effect to name but a few. This requires the building of more and more sophisticated models of an emerging global system in which computers and communication networks are to the twenty-first century what roads were to the first century's Roman empire.

3.4 Solution of Social Problems

Moiseev (1984) says that "past efforts in modeling the socio-economic sphere were largely concerned with the evolution of economic factors. The studies of the Club of Rome offer an example. Yet a purely economic analysis can offer little help in what is in fact most important, namely, the search for ways to resolve the contradictions that are tearing human societies today. The problem of identifying contradictions together with procedures for resolving them through compromises defines the most important branch of research activities today. Methods for finding not merely acceptable compromises but also mutually advantageous compromises may one day exert a decisive influence on the further development of human societies. The theory of compromises is currently one of the rapidly developing branches of science. New approaches and methods have been identified in recent years that make it possible to find mutually advantageous variants of compromises in complex contradictory situations."

The complexity and interrelatedness of the global issues requires technical expertise combined with effective public forum, which is accessible and understandable to a wide spectrum of groups and organizations. The builders and users of models for global issues are geographically scattered, use different languages, reside in different time zones and have different levels of expertise and need for technical details.

Their work to resolve the conflicts based on the quantitative facts and figures of computer simulation requires an asynchronous communication mechanism and can be facilitated by interactive gaming simulation. Consequently, the asynchronous scheduling among dispersed, dissimilar models via global data communication networks becomes a vital tool. The CAC approach allows integrated management of all communication modes: man-to-man, man-machine and machine-to-machine.

3.5 Methodology

The technique described in this paper is a result of the synergism of several recent developments, such as the trend to distributed computing, particularly in the area of databases and knowledge based systems, spread of telecomputing, the developments in computer modeling and simulation and some recent theoretical developments of the scheduling algorithms. We will describe qualitatively a new application of these developments, specially the asynchronous scheduling algorithms, an application of distributed computer simulation system (DCSS) to the field of communication. We call this new area of application of computer simulation technology "Computer Aided Communication" or CAC.

In reference to the International Standards Organization's (ISO)/Open System Interconnection (OSI) reference model, our discussion is confined to the Application Layer (Utsumi, 1982). To outline CAC, we will trace its sources in the recent development of the relevant disciplines.

4. HISTORICAL ROOTS OF CAC AND RELATED FIELDS

The art of model building is ancient and predates the advent of computers. Computer hardware, however, was the first totally "plastic medium" which permitted to model dynamically wide variety of the real processes.

4.1 Interactive Gaming Simulation

Computer models are increasingly used as an advanced design tool. The systems represented by data structures are "animated" and the (simulated) performances of the virtual system are "measured" and evaluated. It is often the "measurement" part of the computer experiment which yields major benefits of the exercise. The variables which are inaccessible or buried in noise tend to stand out clearly and augment the designer's and/or user's understanding of the relevant process performances.

The computer model, when compared to a hardware test bed or prototype, offers a unique advantage as it allows one to develop meaningful statistics of ill defined, stochastic or chaotic systems. The real systems which operate with uncertainty, for example, social systems, can be effectively characterized by this technique. Computer models also aid visualization of the mechanism of the processes which cannot be observed directly and are too complex for unaided imagination.

The interactive mode of computer simulation, which uses animated graphics to feed in the "real time" the effect of players choices, becomes further enhanced in the gaming mode, when several players interact simultaneously with the same simulation model and obtain an immediate feedback not only about simulated part of the system but also about the tactical choices of other players.

4.2 Distributed, Parallel Simulation

The need for more memory, larger processing speed and utilization of other computational, communication and display resources is increasingly more often solved by recourse to the distributed processing. This trend has a specific significance in the field of the real time computer simulation, which is used as a part of controllers which regulate the course of the actual real processes. This is the application in which the computer is indeed operating in analogy with human brain.

4.3 Distributed Modeling

We have explained above why the trend to the distributed processing is particularly urgent in the field of simulation. The motivation is the same as in other areas - the need for faster computing. There is another motivation for the distributed approach to models of social systems which is one of the key concepts behind the CAC technique. To notice the other motivation we must view the phenomenon of the distributed modeling along the whole hardware to software dimension and shift our attention from the computational machines to the methodology of developing models.

The RAND Strategy Assessment Center (Davis, 1983) can serve as an example of the centrally developed distributed model. The hardware is distributed for speed and power, but the software is developed from the unified point of view. The unified systematic approach to the aspect of the reality, which is to be captured by the model, is indeed essential part of the traditional approach to the modeling methodology. The major part of developing computer simulation models deals with the concepts, terminology and selection of the proper sets of equations or non-numerical algorithms. It is this initial phase of the model development, in which the originally diversified views of the contributors are unified into the coherent view of the problem, which is critical for eventual credibility and acceptance of the model.

In this early stage of modeling, "relevant" aspects are being included into the model and some other variables and processes are "neglected." Modeler absorbs the partial views of the assembled teams of experts and must often reconcile conflicting terminologies and conventions of different disciplines. The success of the later model is critically dependent on this early stage, as it determines the acceptance of the model results by the users.

The sensitivity of the model credibility and acceptance to the environment is increasing as we move from hard to soft sciences and is one of the major obstacles to the application of modeling methodology to the solution of social problems, particularly in the situations which include confrontations. The reasons for this barrier are well known, yet persistent.

The CAC aims to circumvent or even, in somewhat paradoxical manner, to exploit this feature of the model building. The CAC techniques can manage and organize communications not only among models but also modelers and users (policy makers), and also among users of different views. The communications include not only text, but also voice (analog or digital) store-and-forward (asynchronous) message exchange, graphics and various video formats, which are used to convey and interpret results of distributed simulation models.

4.4 Need for Interconnections of Dissimilar Models

The computer models are rarely neutral. In the situations where players are to gain or loose, the acceptance of a particular computer model tends to favor one party, usually the one who has developed the model. This is true even in the situation of plus sum game, in which both parties benefit from the adoption of the common strategy. Models utilize certain concepts, certain point of view and good models often contain a philosophy of possible solutions. The negotiation is half won, once one can make the opponent to "see the things our way." An acceptance of our computer model aids a way of achieving that. By the same token, the art of propaganda consists of implanting concepts, slogans and labels, i.e., imposing acceptance of a certain model of reality and of forces affecting it.

These examples illustrate a mechanism which in a less extreme form is present in the development of every computer model, and prevents a development of a large "master model" which could be used to solve social and economical problems. The uneasy feeling, caused by the overtones of an Orwellian society, controlled by an impersonal elite, operating and perhaps controlling a huge supercomputer containing vast databases concerning public and perhaps even private affairs, is another example of the same mechanism which operates in building, propagation and use of the models in the social arena. We all, being players in the public affair games, are concerned about the model which may be accepted or imposed on us and may be used to allocate the winnings and losses and in selection of the common policies and strategies.

This non-neutrality of the model and language, which operates generally in human affairs, can be seen as an analogy or perhaps a complement of Heisenberg's Uncertainty Principle: by observing the system, we affect and change it. Here, merely by describing the system, we change our perception of that system.

While this mechanism operates on all levels, it becomes particularly apparent in the building of computer models based on soft rather than hard sciences, applied to social problems and involving resource allocation issues. It is in the marginal cases, where these effects may be weak and unsuspected, that neglect of this mechanism is likely to cause problems. These problems are often manifested merely by nonacceptance of the model, sometime dubbed as NIH (=Not Invented Here) syndrome and not always recognized as major obstacle to application of simulation to the social issues.

4.5 Interconnection of Distributed Databases

The same mechanism operates on the databases, again both computerized or not. The databases are more widespread than computer models and so the operation of this "Uncertainty Principle for application software" is better known; it is centered around the concept of the "access." As the technology is supplying the computational resources in ever increasing quantities, hardware aspect is becoming less important. The other motivation for the move to distributed database is not hardware driven but related to the application and use of the data. The users of the databases are scattered geographically and those who generate, collect or create the data may need to exercise a measure of control on their data (e.g., restrictions of trans-national data flow recently emerged in European countries (Utsumi, (1978), Norman, (1981)).

To accommodate this need for local control of the data and some transactions, database programs allow some users, "owners of the data," to issue specific permits. Resulting set of "permits," which gives different privileges to different users, reflects and modifies the power structure of the organization which is developing the database. As with computer models, when the mechanism of the extended Uncertainty Principle is ignored, the resulting database is abandoned as inflexible, impractical or irrelevant.

Distributed databases exhibit new complexity which is directly related to the fact that multiple users or players with different payoffs in the game are using the same set of data. This new complexity includes the traditional aspects, such as issues of integrity, concurrency, location and duplication of the data. However, the aspect of access, and control of data are becoming dominant issues.

Next generation of the databases combined with simulation models will expand considerably the ability of game players to assign dynamically wide variety of constraints and permits, which will allow information to flow freely through the network, be modified as needed, accounted for, edited as needed, delivered to proper audiences, protected, rewarded, etc.

The aspects of merging dissimilar databases and the social or political concerns, such as concerns about the inevitably creeping motion to interconnecting databases developed by governmental agencies, all those aspects closely parallel the concerns and problems which we have discussed in connection with computer models.

4.6 Integration of Simulation Model and Database

This similarity between the two fields of telecomputing is not surprising when viewed in the light of the latest developments in the techniques of the simulation. The area of simulation, databases and of artificial intelligence are converging. Large scale state-of-the-art simulation projects are heavily database based and often use commercial relational databases, such as ORACLE or INGRES. The database is used to store initial data and parameters as results of the simulation. Manifestation of the same trend are commercial simulation packages, such as TESS by Pritsker & Associates, which combine graphics and database with simulation.

As the speed of computation will increase, the relational databases will be used concurrently with the simulation. The database will be able to display any object or particular attributes of an object or entity and a model running in the background which will be able to animate the object, and so convey the data and change of the data/objects with a particular scenario or parameters of the simulation run.

4.7 Advanced Programming Languages

One of the outstanding fifth generation programming languages is MODEL developed by Prywes and others at University of Pennsylvania (Cheng, et al (1984), Prywes, et al (No Date and 1985) and Tseng, et al (No Date)). MODEL is a powerful, non-procedural computer language which generates programs automatically, using as input a description of the problem rather than a description of the solution. It allows sequential or parallel processing on dispersed, dissimilar mainframe computers (currently IBM and DEC/VAX). The detailed design of communications and synchronization is performed by the automatic programming system.

With MODEL, novices can solve problems that once could tax even the most skilled programmer. To communicate a problem to MODEL, it is only need to specify the problem mathematically, without regard to how it will be implemented. Because only a basic knowledge of mathematics is required, any nonprogrammer still feel comfortable using the system. Compared with COBOL, MODEL uses only one-fifth as many statements.

MODEL can support bottom-up approach to building a large scale system from existing subsystems, as well as the conventional top-down approach. The MODEL language specification is easy to modify. It is incremental, in the sense that variables or equations can be added at the end or in any place, since order of statements is arbitrary.

Multiple designs are examined automatically. The computer chooses the most efficient. Operations are sequenced automatically, and MODEL automatically checks for consistency and completeness. Job control language to set up and execute concurrent processing is also generated automatically. It generates command language programs that schedule the execution of modules, maximize parallelism, and set up the communications among modules which will be executed in parallel. A component graph is constructed which serves as a basis for scheduling. This graph consists of nodes representing parallel components and edges indicating sequential or arbitrary order of initiation of these components.

Entire independently developed systems may be easily connected. Thus, the creation of a new system that encompasses old system would not require design of a new system, only at most additions of modules to convert data. In the cooperative computation mode, individuals or groups develop systems independently, motivated by their own interests. They may later discover that by joining their systems they can have even greater capabilities than the total of the separate systems. All that is required then is to develop modules that convert the data to a common format and form a new interconƒnected system.

The MODEL language has been used to make possible the distributed processing of Project LINK which is an econometric forecasting simulation of world economy with individual country's economic model. Project LINK was originated and developed by Nobel Laureate Professor Lawrence Klein of University of Pennsylvania. Professor Prywes plans to locate the distributed processors in Japan and West Germany which will be interconnected by global VANs. We are assisting him to realize his plan with the use of the extended line of the U.S. CSNET to Japan.

Another relevant trend is the use of the object oriented programming languages (Cox, 1984), such as SMALLTALK or PROLOG, which work with sophisticated data structures. Objects can be data entities describing the real, physical objects, they may correspond to files or data sets, tables in the database or they may represent executable programs.

When we view this modern trend to object oriented programming in the above sketched environment of the distributed computer simulation systems (DCSS) communicating over the global data networks, we see the objects being sent from one location to another, executed and animated as a part of a simulation run, used as data envelopes and equipped with a set of attributes which define their access status, sensitivity of the data and so on.

The collection of the object traveling on loosely connected networks of the dissimilar computers forms a meta-language which is able to transcend the barrier to the application of the computer models to the social problems described above. The language, which is formed from independently defined object/concepts which the players can adopt or bypass, does not impose the specific world view of a language designer on the players. In this dynamic aspect which allows the different groups to adopt, drop and modify the objects used for mutual communication, the meta-language composed of the objects resembles the natural languages and shares some of their efficiencies. However, it differs from the natural languages as it consists of the objects which conform to certain rules of the formal syntax and can be directly executed or read by all component computer models in the system.

4.8 Summary

It is the hardware aspect of the "distributed" system which makes for the quantum leap in the use of the computer systems to the social issues - as it allows them to tailor the systems to the existing relationships and power structure of the society. As the society is not a monolith, controlled and manipulated form one center, a single large supercomputer operated by an elite team of experts does not satisfy criteria enumerated above, for an environment to be able to produce a successful computer model and simulation. Below, we will describe the methodology for building of the large scale models of the social systems which extend the benefits of the distributed simulation into the software aspect of the problem.

5. DEVELOPMENT OF ASYNCHRONOUS MODELING

On the boundaries of recent developments of various disciplines described in the preceding sections, one can discern a new application of the computer simulation technology, which will be the communication intensive discipline. Affected fields are education, large scale industrial and scientific projects and management of multinational corporations. However, the application which appears to have the most potential to benefit from this new technology is the global resource allocation, and environmental and ecological management.

These applications represent the "social order" which is a prerequisite to development of any new technology. Modeling of energy and other global resources was discussed under the key word of GLOSAS Project, which stands for GLObal Systems Analysis and Simulation (Utsumi, 1972 to 1985), to which readers are referred for further details. A comprehensive and coherent model of the global resources is a prerequisite to consensus on proper resource allocation and ecologically sound policies.

Allocation of natural resources on merely a national scale is a complex problem, requiring skillful blending of the technical expertise, political consensus and complex administrative and financial systems. On the global scale, the communication problems so far prevented all but the most rudimentary manifestation of the emerging consensus - that something more than business as usual is to be done soon, to counteract and reverse destabilizing effects of the increasing population to the natural resources ratio and attendant increase in international tension.

We have summarized above the obstacles which prevent the successful development of the large scale, comprehensive computer models in those areas. We not only do not have consensus, but we do not even have a common picture of what the problem is. The simulation is the tool for the development of all that; first - of the common picture of the problem, then of the common terminology and gamut of possible strategies and finally the development of the consensus on global strategies. The interactive gaming simulation tests the mutual interaction of strategies of the players in the game and allows assessment of future consequences of the individual local strategies.

The first step in the evolution of the distributed gaming simulation, which may be the implementation test bed for Computer Aided Communication, is the development of the isolated, individual models of the reality. Many such models already exist, ranging from economical models, e.g., Meadows' "Limit to Growth" (Meadows, 1972 and 1974) and Onishi's FUGI model (1983 to 1984) to purely technical ones - models allowing engineers to design a car engine or allowing meteorologist to predict weather.

Incidentally, Onishi's FUGI model with economic and resource/energy forecasting submodels of over 125 individual countries was once used by economists of India, New Zealand, Australia and the United States with the HUB computer conferencing system of the Institute For The Future (IFTF) in Menlo Park, CA, over global VANs when East-West Center in Honolulu conducted an international economic affair gaming simulation, though the FUGI model was processed by a central mainframe computer at that time (Lipinski, 1982). The GLOSAS Project will attempt to distribute FUGI's national economic and resource models to as many countries as possible, starting between the U.S. and Japan with the use of the extended line of the U.S. CSNET to Japan.

Each such model represents an aspect of the multi-faceted reality, and moreover it tends to contain a bias of the model authors. The bias tends to be more important for the models dealing with the soft sciences and issues and may be decisive feature for models dealing with the natural resource allocation. Each such model is developed freely, without a need for a common world view, or even a same methodology, simulation language or similar hardware.

The conventional approach to the modeling is to obtain and consolidate common views and concepts. In contrast to that, GLOSAS Project and CAC admit and acknowledge persistence of conflicting opinions existing in the real world, and hence concentrate on the interconnection of dissimilar models.

The second step is to create the globally distributed computer simulation system (DCSS) by the interconnection of independently created and verified individual self-standing models. This second task is being addressed by the standards for the communication technology manifested by the current ISO/OSI standard and concurrent growth and spread of the computer networks. Additional standards, subdividing the 7th OSI layer for application programs, will cover the specifics of the Open Model Interconnection (OMI) and allow exchange of the model specific messages and objects.

The third step is the most interesting: What have we achieved if we interconnect dissimilar computer models, developed by experts of different disciplines located at different sites, using different methodologies, financed by different institutions, which themselves are tied to opposing power structures with conflicting interests? Imagine that we have done that: We have a public database of the existing computer models which are conditionally available for concurrent asynchronous execution. The database lists the assumptions of the models, their inputs and output variables, network addresses, mode of access, etc. Let us call the resulting Tower of Babel of inconsistent and contradictory models an Incoherent Modeling Network.

The interconnection of the heterogeneous models into one communication network creates a distributed virtual process, which can serve as semantic laboratory, a test bed for development of the computer tools for interpreting meaning of the I/O files of one model in terms of another model. In the process of such interpretation/translation, the meaningless incoherent messages, some of them at any rate, have a chance of becoming information.

The meaning of the message depends on two complementary structures: the physical structure of the message itself, the sequence of the symbols, of letters and words. The meaning also depends on the model of the world which the recipient of the message is running. If the recipient, be it man, organization or machine, is not running a model of the world which can interpret the message, then there is no meaning, hence no information transfer. (The former corresponds to our different natural languages and the latter to cultures.) The completion of the third step will not in itself create any new meaning, consensus or understanding on the "airwaves" - it will merely put the incoherent messages into a computer-readable form, so to speak.

The fourth step is devoted to the task of translation and correlation. There is a large selection of techniques available, which allow one to find and extract relevant information from the usually vast number of irrelevant messages which we call noise. These techniques range from the mundane regression analysis which calculates the positive or negative correlation between two variables, to the sophisticated techniques, such as the Cluster Analysis used in the Mathematical Taxonomy.

The task of the translation in our CAC is not accomplished by mechanical search for the statistical correlations between the parameters and outputs of the various models, even though, in some cases, the techniques of the artificial intelligence can be used to detect unsuspected relationships in classes of corresponding models. An example of such models may be classes of the economical models which simulate national economies, industries and various technical aspects, such as material or financial management, transportation, etc.

Prerequisite for interfacing the content of models is interpretation of elementary dimensions of the models. In our example of the economical models, the dimensions would include the virtual time in the simulated universe and the geography which defines the boundaries of the different economies, etc.; Dimensions facilitate the interpretation of the results of one model in terms of the inputs of the other models. These meta-models, or translating models, can be seen as new abstract concepts, which can only emerge in a meaningful fashion after the more concrete concepts are well anchored in the actual proxies.

The building of the interconnecting and interpreting models will be analogous to the creative capability which can suddenly see, in an intuitive flash, a connection of previously unrelated thought processes. This specialized, high-level translation process will be complemented by the process more akin to the usual, present day translation. In the process of executing the individual models, it will become possible to build dictionaries which can relate the concepts of some models to those of the other ƒmodels.

In contrast to the usual dictionaries of common languages, these dictionaries can be quantitative, dynamic and will have full benefit arising from the semantic clarity which the concepts used as basis of a computer model tend to acquire. This clarity of concepts, which the users of good computer models tend to acquire, represents the semantic benefit of the simulation.

6. MECHANISM OF THE ASYNCHRONOUS, DISTRIBUTED SIMULATION

In the rest of this text we draw a loose, yet exact analogy between the simulation runs of the global modeling network and the transactions of a conventional Discrete Event (DE) simulation run. The most familiar form of the computer simulation is the Continuous System (CS) modeling. CS models essentially solve a system of the differential equations, which use time as the independent variable. Essential feature of the CS models is the time loop, which increment the time variable in the small steps and updates the state variables in the model. In the DE model, time is a dependent variable. The model is represented as a series of events which have specific causal and temporal relationship.

The essential feature of the DE simulation is the Scheduling Algorithm, which increments the time in finite chunks and schedules the events in proper sequence. The computational load of the DE simulation thus consists of the isolated sets of the operations, i.e., transactions which can in many cases be carried over in parallel.