Chapter 30 Governing Social-Ecological Systems

Abstract

Social-ecological systems are complex adaptive systems where social and biophysical agents are interacting at multiple temporal and spatial scales. The main challenge for the study of governance of social-ecological systems is improving our understanding of the conditions under which cooperative solutions are sustained, how social actors can make robust decisions in the face of uncertainty and how the topology of interactions between social and biophysical actors affect governance. We review the contributions of agent-based modeling to these challenges for theoretical studies, studies which combines models with laboratory experiments and applications of practical case studies.

Empirical studies from laboratory experiments and field work have challenged the predictions of the conventional model of the selfish rational agent for common pool resources and public-good games. Agent-based models have been used to test alternative models of decision-making which are more in line with the empirical record. Those models include bounded rationality, other regarding preferences and heterogeneity among the attributes of agents. Uncertainty and incomplete knowledge are directly related to the study of governance of social-ecological systems. Agent-based models have been developed to explore the consequences of incomplete knowledge and to identify adaptive responses that limited the undesirable consequences of uncertainties. Finally, the studies on the topology of agent interactions mainly focus on land use change, in which models of decision-making are combined with geographical information systems.

Conventional approaches in environmental economics do not explicitly include non-convex dynamics of ecosystems, non-random interactions of agents, incomplete understanding, and empirically based models of behavior in collective action. Although agent-based modeling for social-ecological systems is in its infancy, it addresses the above features explicitly and is therefore potentially useful to address the current challenges in the study of governance of social-ecological systems.

Introduction

For millennia, human activities have affected their environment. In ancient times, the use of fire and tools enabled humans to learn to live outside their original environment—the savannah of eastern Africa. The development of agriculture about ten thousand years ago, and industrialization during the last two hundred years, have generated massive population increases and intense uses of natural resources. Now, we live on a human-dominated planet. Human activities have transformed the land surface, altered the major biogeochemical cycles, and added or removed species in most of Earth's ecosystems (Vitousek et al., 1997).

This chapter reviews the efforts by many scholars to use agent-based computational models to study the governance of social-ecological systems. This field is truly interdisciplinary. It will be difficult, if not impossible, therefore to restrict our focus solely to economics. Although economics will be our starting point, we will include studies from other disciplines. To facilitate communication across disciplines we will use an organizing framework in the second section of this chapter. To structure our chapter, we identify three main challenges for the study of the interactions between human activities and ecosystems.

• What conditions enhance the likelihood of cooperative solutions to the massive number of social dilemmas that confront social-ecological systems? This relates to the problem of preventing overharvesting of common-pool resources such as fish stocks, forests, and fresh water.

• How do economic agents make effective and robust decisions given the fundamental uncertainty of the complex dynamics of the social-ecological system?

• How can the topology of interactions among actors be explicitly included in the analysis of the first two questions given the importance of interactions to an understanding of natural resource dynamics?

The aim of this chapter is to show the contribution of agent-based computational economics to these challenges. We emphasize the linkages between field research, laboratory experiments, and agent-based modeling. Pure analytical models have proved to be essential tools for analyzing highly competitive markets and other settings with strong selection pressures (Ruttan, 2003). When trying to understand how and why individuals engage in collective action, however, analytical models have not proved as useful. In the field and in the experimental laboratory, we have observed many settings in which individuals overcome the incentives to free ride, increase the levels of inter-personal trust, produce public goods, and manage common-pool resources sustainably (Bromley et al., 1992, Gibson et al., 2000a, National Research Council, 2002, Ostrom and Walker, 2003, Dietz et al., 2003). Candidate theories for explaining these surprising empirical results are too complex to be usefully pursued using only analytical techniques. To understand these phenomena agent-based modeling has become an essential tool complementing empirical methods. Other chapters in this volume (Brenner, 2006, Duffy, 2006) also address the combination of laboratory experiments and agent-based modeling. Their contribution focuses more on learning models, while our focus is on public goods and common-pool resource experiments using several models of human decision-making. It is important to realize that every method used to study social-ecological systems has its methodological problems. We will therefore emphasis in this chapter the plurality of approaches, which may unravel the complexity of the systems when findings are consistent with all the types of approaches used.

The studies reviewed in this chapter differ from those most frequently addressed by environmental economists. Conventional economic theory predicts that when agents have free access to a common-pool resource they will consume ecosystem services to the point where private costs equal the benefits, whereas externalities are imposed on the rest of the community. This can lead to the well-known tragedy of the commons (Hardin, 1968). Traditionally, economists study the management of ecosystems in terms of harvesting ecosystem services from renewable resources. Substantial progress has been made during the last 30 years. Prior to 1970, models were mainly static, such as the seminal work on renewable resource harvesting by Gordon (1954). During the 1970s, the trend shifted toward dynamic systems for the economics of renewable resources. The resulting optimization problem was addressed by dynamic programming, game theory, and equilibrium analysis (Clark, 1990, Dasgupta and Heal, 1979, Mäler, 1974). Irreversibility and uncertainty have been addressed since the early 1970s (Arrow and Fisher, 1974, Henry, 1974) and remain among the main foci of environmental economics (e.g., Chichilnisky, 2000). Recently, economists have started to include non-convexities of ecosystems into their analysis of optimal management of ecosystems (Dasgupta and Mäler, 2003, Janssen et al., 2004).

In simple models in mainstream environmental economics, a representative agent is presumed to have perfect knowledge (or knowledge on the probabilities of outcomes) and to maximize utility of consumption for an infinite time horizon. Such an approach results in interesting insights. Representing agents as maximizing known utility functions is, however, of limited use when systems are characterized by non-convex dynamics, structural uncertainty, heterogeneity among agents, multi-attribute utility, and spatial heterogeneity. Evidence is accumulating that social-ecological systems frequently do have complex, non-linear dynamics. This affects the type of governance that may lead to sustainable outcomes (Scheffer et al., 2001). Initial steps has been taken to include such non-linear dynamics in environmental economics (Dasgupta and Mäler, 2003). Furthermore, increasing evidence exists that agents are able to self-govern some types of common-pool resources without external governmental intervention but do not always succeed (Bromley et al., 1992, Ostrom, 1990, National Research Council, 2002, Ostrom et al., 1994). The question is how to analyze ecosystem management problems with spatially explicit, non-convex dynamics influenced by multiple stakeholders with divergent interests and who consume different types of ecosystem services. We need new tools. Agent-based modeling is a promising tool for the analysis of these complex problems (Janssen, 2002a).

Several developments outside environmental economics during the last thirty years have influenced the current state of agent-based modeling of social-ecological systems. We will briefly discuss some of these developments. Since the early 1970s, scholars from system dynamics have developed and used integrated models of humans and their environment (Ford, 1999). Prime examples are the World 2 and 3 models of Forrester (1971) and Meadows et al., 1972, Meadows and Behrens, 1974. The World 2 and 3 models simulated the long-term interactions between population, industrial and agricultural production, resource use, pollution and food supply at an aggregated global level. A core finding was that continuing early 1970s' trends would lead to an overshoot and collapse in terms of population and economic development. The World 2 and 3 models were highly criticized for the subjectivity of the assumptions and the lack of rationality of the decision-making actors within the model (Cole et al., 1973, Nordhaus, 1973). In fact, the actors, economic sectors on a global level, reacted in a predetermined way.

The first type of agent-based model for governing social-ecological systems that we were able to trace in the literature is Bossel and Strobel (1978). They developed a model to address two lacunae in the World 2 and 3 models—namely, their failure to account for cognitive processes and their usual neglect of normative criteria and changes in these criteria. In fact, the Bossel and Strobel model is of a cognitive agent interacting with the global system. Their agent bases its decisions on the state of the global system, using indicators, so-called system's orientors, like existence needs, security, freedom of action, adaptivity, and effectiveness. This agent receives information about the state of the system and decides to change priorities or aspirations, which affect the investment decisions of the agent. Inclusion of these “intelligent” agents prevents the preprogrammed “pollution crisis” from occurring. It also leads to policies producing very satisfactory overall results, provided the planning horizon and the control sensitivity are sufficiently large. The current field of integrated modeling of humans and the environment still faces similar problems, uncertainty, subjective assumptions and lack of behavioral models, to those of the initial models (Janssen and de Vries, 1999). Core questions remain regarding how to deal with uncertainty and subjective assumptions and how to include human dimensions.

Another field that contributed to the development of agent-based modeling of social-ecological systems is individual-based modeling in ecology, which really took off in the late 1980s (Huston et al., 1988). Individual-based modeling refers to simulation models that treat individuals as unique and discrete entities who have at least one property, in addition to age, that changes during the life cycle, e.g. weight, rank in a social hierarchy, etc. Often motivated by pragmatic reasons, individual-based models are used to study systematically the behavior of organisms in complex (spatially explicit) environments (Grimm, 1999).

In the artificial intelligence field since the late 1980s, scholars developed tools for natural resource management (Coulson et al., 1987). Well known are geographic information systems and expert systems, but also a number of models have been developed that included intelligent agents interacting with their complex environment (Anderson and Evans, 1994). An interesting early example is the PHOENIX model on fire management (Cohen et al., 1989). The model simulates a forest fire and the actions of intelligent agents, representing bulldozers and airplanes. The model is an event-driven simulation model, meaning that the agents perform real-time tasks based on events that happen in their local environment. Every five simulated minutes of the model, the agents are synchronized to allow coordination among the agents. The model is aimed at evaluating fire-fighting plans in various scenarios.

Bousquet et al. (1994) developed an objected-oriented model of natural resource management of fisheries in the central Niger delta. Based on fieldwork, an artificial world was created where different scenarios of rules of when and where to fish in a wetland area were analyzed for this impact on long term viability of the natural resources. The existence of space-sharing rules was found to be essential to avoid overfishing.

Deadman and Gimblett (1994) constructed a system that handles the complexity of goal-oriented autonomous human agents seeking recreational opportunities in natural environments. The model simulates the behavior of three types of visitors and their interactions in an event-driven GIS environment of a park environment using intelligent agents: hikers; bikers; and visitors transported in tour vehicles. The results of hiker interactions with other users have been used to provide feedback about the implications for alternative recreation management planning.

Complexity science is still another foundation for the study of the governance of complex social-ecological systems. Social-ecological systems can be viewed as complex adaptive systems—systems in which the components, and the structure of interactions between the components, adapt over time to internal and external disturbances (Holland, 1992a). Order in complex systems is emergent as opposed to predetermined. The system's history is irreversible, and future behavior is path dependant. The system's future is often unpredictable due to the non-linearity of many basic causal relationships. The variables that affect performance are both fast and slow moving. If information about slow-moving variables is not recorded for a long period of time, substantial surprises can occur when a slow-moving variable reaches some threshold. In social-ecological systems, the key components are individuals and institutions. With institutions we refer to the formal and informal rules that shape human interactions. Individuals may change their relations with other individuals, their strategies, and the rules they are using. In fact, individual strategies and institutional rules interact and co-evolve, frequently in unpredictable ways. For example, the peasants who were starting to drain the peat mires on a local level more than 1000 years ago in the precursor of the Netherlands did not foresee the large-scale consequences in the few hundred years on the larger-scale landscape (lowering of the surface by about 2 cm a year), leading to new institutions (like waterboards), and different practices (livestock instead of agriculture).

From this perspective, the question arises of how to govern social-ecological systems. In systems that are indeed complex, one needs to understand processes of organization and reorganization including collapse and the likely processes that happen after collapse. Does a system have one and only one equilibrium to which it returns after a major shock and temporary collapse? Are there multiple equilibria with different characteristics? How easy is it for a system to flip from a desirable equilibrium to an undesirable one? These are crucial questions.

The complex adaptive systems perspective provides us the view of individuals within a variety of situations structured by the biophysical world, the institutional rules, and the community in which they interact. Within ongoing structures, individuals search out perceived advantageous strategies given the set of costs and benefits that exist and the strategies that others adopt. Boundedly rational individuals trying to do as well as they can in uncertain situations continuously tinker with their strategies, including trying to change the rules that affect particular situations. They may look for loopholes in the law, particularly if they think others are doing the same. They may check out the level of enforcement by occasionally breaking rules. Those who have responsibility for changing the rules of an institution also experiment with new rules and try to learn from others why other institutional arrangements appear to work better than their own.

Agent-based models are a suitable methodology to study these complex social-ecological systems in a formal manner for the following reasons:

• Agent decisions are based on internal decision rules; this fits very well with the increasing insights from experimental social science that humans use various types of heuristics in different situations (Gigerenzer et al., 1999, Gigerenzer and Selten, 2001).

• The explicit inclusion of agent interactions helps to integrate the increasing insight of the importance of communication in managing social dilemmas (Ostrom et al., 1994, Ahn et al., 2003, Ahn et al., 2004).

• Agent-based modeling shares similarities with models used in ecology, such as individual-based models, system theory, and the inclusion of space. Therefore, agent-based modeling facilitates collaborative efforts of ecologists and social scientists.

• Agent-based models are suitable for modeling complex adaptive systems, in which the interactions of individual units lead to larger-scale phenomena.

• Agent-based modeling makes it possible to address the problem of scale explicitly (Gibson et al., 2000b).

The perspective of social-ecological systems as complex adaptive systems provides us a useful stepping stone for using agent-based modeling for the study of social-ecological systems. In the next section we discuss a general framework of social-ecological systems that we will use as a guideline to discuss the work done in this field.