The Role of Intelligent Agents in Integrated Environmental Management

Sita Smith and William Mackaness
Department of Geography, University of Edinburgh, Edinburgh EH8 9XP, United Kingdom

Computer systems play an ever increasing role in environmental management. From data collection and analysis, through to its visualization (typically manifest in the form of Geographic Information Systems). The effective use of such technology is hampered by the complex interface, numerous permutations in the composition of spatial data and the poor feedback during the various stages of processing. This paper suggests that intelligent agents can be used to facilitate the mechanics of decision making, especially where decisions are 'shared' among the system and a range of stakeholders. This extended abstract begins with a description of the nature of environmental management and intelligent agents and how the latter can be used to support the (typically complex) spatial decision making involved in environmental management. It then discusses a potential implementation strategy.

What is environmental management?
Data for environmental resource management is multi-disciplinary, spatial and varied in nature. In developing countries, where most sustainable resource management projects are funded through external loan agencies, decision makers are faced with data that varies both in quality and quantity. Under these circumstances, it is extremely difficult for decision makers to take into account all the information necessary for the trade-offs and compromises essential to developing and choosing management alternatives. In general, resource management projects are selected within a framework (the project selection cycle), that must make explicit the interaction between analysis, planning and management, but also provide guidance on the sequence of these activities. In the early stages of the project selection cycle, there is an imperative to develop feasible, politically and economically acceptable solutions for a variety of special interest groups.

Why does it require decision support?
The process demands both scientific input as well as input from social, political and economic domains. Such expertise is often unavailable at the right time, place or price. Decision makers working towards a mutually acceptable resource plan need assistance in developing, representing and analysing their decision space. Given the variability of data sources and dearth of appropriate expertise at the appropriate times, the use of computerised decision support systems becomes significant (Charleboisle et al., 1995). Decision support systems are computer-based systems that help users deal with problems, through data and models, within a formal framework (Fedra, 1995). They are a tool to maximise the effectiveness of a decision maker's cognitive processes (Zachary, 1994). In the context of environmental project selection, a decision support system would offer the user access to the datasets associated with the problem, an environment in which to model the problem, and analytical tools to evaluate solutions and strategies. Unfortunately, the complexity and difficulty of environmental management problems is reflected in the heterogeneity of the tools - both microeconomic and environmental - that are required to deal with them. Environmental decision support systems may include GIS, spreadsheets, modelling tools and knowledge-based systems (Lam & Swayne, 1991). A consequence is that even in the relatively rare cases of successful implementation, the decision support systems are underused. Tools based on artificial intelligence techniques, such as intelligent agents, are a possible solution.

Intelligent agents
Intelligent agents are computational systems that inhabit dynamic, unpredictable environments (Wooldridge & Jennings, 1995). Agents are self-contained, rational, problem-solving entities, that act to accomplish tasks on behalf of a user (Shoham, 1990). They interpret events in their environment, and execute motor commands that produce effects in that environment, in order to satisfy predefined goals. Unlike other artificial intelligence tools, they have been found to be effective in managing systems that are open, unpredictable, complex and large - all characteristics typical of the environmental management domain.

Agents seem designed to work with the complex interaction dynamics inherent in environmental decision making. Agents represent an interface between the problem, the decision support system, and the user (Chigmell et al., 1994). They are a solution, whereby a non-expert user can navigate both the data sources and structure of the problem as well as the decision support software developed to help resolve it, thus removing some of the cognitive burden from the user's shoulders (Chin, 1990).

Case study
The pilot study for the proposed research revolves around the Segara Anakan estuary in Java, Indonesia. The Citanduy River drains into the estuary, on the South Java Sea. Partly as a result of upstream development, the Segara Anakan is filling up with sediment; it is estimated that the lagoon will be silted up by the end of this century. As the lagoon silts up, waterways become narrower, drainage more difficult, flooding more frequent, fishery production limited, and mangrove forests threatened. The Indonesian Government has invested millions of dollars (ADB Report, 1994) in an attempt to define a feasible solution for the estuary, and to identify development projects for the area that will not exacerbate the problem.

There are a host of socioeconomic pressures on the estuarine ecosystem, including population expansion, tourism, and aquaculture/agriculture development. These are reflected in the vast range of actors (government officials, donor agencies, regional politicians) who have a stake in the project selection process for the development of the Segara Anakan. In this context, a spatial decision support system would have to garner the responses and political stances of these stakeholders and be able to filter out key issues. It is intended to construct a decision support system, with intelligent agent interface, to help evaluate the feasibility of proposed projects in a lagoon/mangrove ecosystem in Indonesia.

Agent-based system development involves building self-contained components that can interact flexibly with other, similar components, towards solving a common goal. A logical first step is to build a framework that incorporates the project selection cycle as well as the knowledge bases and analytical tools on which the project selection process is based (Le Fur, 1995). This would include an integrated decision support system with GIS, a dynamic environmental model, databases and analytical tools. Critically, he focus of this collection of disparate systems would be an intelligent user interface that could handle the goals, tasks and mechanics of the agency. Instead of using the decision support software to handle the problem, the user lets the agent schedule and run the components of the decision support system (Saarenmaa, 1994).

Such a decision support system will, at the very least, ensure that all, important factors are taken into consideration when project feasibility is being evaluated, and at best, be a powerful tool for decision makers.

Asian Development Bank. 1994. Segara Anakan Conservation and Development Report. ECI Ltd, TA No. 1624-INO, Consultant Services Contract No. CPCS/92-156.

Charleboisle, D., Matwinl, S. and Goodenough, D. 1995. Planning with Agents in Intelligent Data Management for Forestry. Proceedings Artificial Intelligence and the Environment Workshop, IJCAI-95, Montreal.

Chigmell, M.H. and Hancock, P.A. 1994. Intelligent Interface Design. In Helander, M. (ed) Handbook of Human-Computer Interaction. Elsevier Science B. V.

Chin, D. 1990. Intelligent Interfaces as Agents, Report ICS-TR 90-06, University of Hawaii Technical Report Series.

Fedra, K. 1995. "Decision Support for Natural Resources Management: Models, GIS and Expert Systems". Artificial Intelligence Applications, 9(3), 3-19.

Lam, D.C.L. and Swayne, D.A. 1991. "Integrating Database, Spreadsheet, Graphics, GIS, Statistics, Simulation Models and Expert Systems: Experiences with the RAISON System on Microcomputers". In Loucks, D.P and da Costa, J.R. (eds) NATO ASI Series - Vol. G.26. Berlin Heidelberg: Springer-Verlag .

Le Fur, J. 1995. "Modeling Adaptive Fishery Activities Facing Fluctuating Environments: An AI Approach", AI Applications, 9(1):85-97.

Saarenamaa H., Pertunnen, J., Vakeva, J. and Nikula, A. 1994. "Object-Oriented Modeling of Tasks and Agents in Integrated Forest Health Management". AI Applications, 8(1), 43-56.

Shoham Y. 1990. Agent-Oriented Programming. Report No. Stan-CS-90-1335, Stanford University, Dept. of Computer Science.

Wooldridge, M. and Jennings, N. 1995. "Agent Theories, Architectures and Languages: a Survey". In Wooldridge, M. and Jennings, N. (eds) Intelligent Agents - Theories, Architectures and Languages - Lecture Notes in Artificial Intelligence Vol. 890. Springer-Verlag

Zachary, W. W. 1994. "Decision Support Systems: Designing to Extend the Cognitive Limits". In Helander, M. (ed) Handbook of Human-Computer Interaction. Elsevier Science B. V.