Ecological models based on people’s knowledge: a multi-step fuzzy cognitive mapping approach

https://doi.org/10.1016/j.ecolmodel.2003.10.027Get rights and content

Abstract

Many types of ecological or environmental problems would benefit from models based on people’s knowledge. To create ecological models with both expert and local people’s knowledge, a multi-step fuzzy cognitive mapping approach is proposed. A cognitive map can be made of almost any system or problem. Cognitive maps are qualitative models of a system, consisting of variables and the causal relationships between those variables. We describe how our cognitive mapping research has been used in real environmental management applications. This research includes examining the perceptions of different stakeholders in an environmental conflict, obtaining the perceptions of different stakeholders to facilitate the development of participatory environmental management plans, and determining the wants and desires for resettlement of people displaced by a large scale dam project. Based on our research, which involved six separate studies, we have found that interviewees complete their cognitive maps in 40–90 min on average. These maps contain an average of 23±2 S.D. variables with 37±3 S.D. connections. People generally put more forcing functions into their maps than utility variables. Fuzzy cognitive mapping offers many advantages for ecological modeling including the ability to include abstract and aggregate variables in models, the ability to model relationships which are not known with certainty, the ability to model complex relationships which are full of feedback loops, and the ease and speed of obtaining and combining different knowledge sources and of running different policy options.

Introduction

Modeling ecological or environmental problems is a challenge when humans are involved. We identified four types of problems where gaining insights or predicting system behavior can be very difficult. The first type involves human behavior and how human actions can affect an ecosystem. Examples include management of sport or commercial fisheries, where anglers and fishing effort can have a considerable effect on the ecosystem. The behavior of commercial fisherman and anglers needs to be accounted for in these models because knowledge of fishing effort in response to regulations is needed to choose appropriate management options (Dreyfus-Leon and Kleiber, 2001, Radomski and Goeman, 1996). In such cases a modeling tool to determine perceptions of fishermen concerning a fishery and their probable actions given different management scenarios would be useful.

The second type of problem involves instances where detailed scientific data are lacking but local knowledge of people adapted to an ecosystem is available. Much could be learned from incorporating this local or indigenous knowledge but typically models have no means to do this.

The third type is “wicked” environmental problems (Mason and Mitroff, 1981). These problems are complex, involve many parties, and have no easy solutions or right answers. However, decisions must be made. A useful modeling tool for analyzing such problems would bring together the knowledge of many different experts from different disciplines, be able to compare their perceptions and to simulate different policy options, allowing for discussion and insight into the advantages and disadvantages of possible decisions.

Finally the fourth type of problem concerns ecosystem management where public involvement is desired or even mandated by law. Most ecosystem models typically address things such as primary production while the public is concerned with properties such as human health or costs of different management options (Hobbs et al., 2002). In these cases the modeling method should be able to incorporate public opinions about these higher-level variables of concern to the public. In addition, the model could be used to inform the public regarding different management options, and enable public support for management decisions.

All of these types of problems would benefit from models based on people’s knowledge. Cognitive mapping provides a means to do this. Özesmi (1999a) developed a multi-step fuzzy cognitive mapping approach for analyzing how people perceive an ecosystem, and for comparing and contrasting the perceptions of different people or groups of stakeholders. The multi-step approach is a synthesis of relevant useful techniques from many disparate disciplines on cognitive mapping. This article describes this multi-step approach in detail, providing a methodological tutorial and some examples from our own research. In addition, to our knowledge this article is the first review of cognitive mapping for ecological modeling and environmental management. First we give a brief introduction to cognitive mapping including its history and some examples of its uses. We compare FCM to some other techniques and discuss the advantages and disadvantages of cognitive mapping. Based on our research and examples from the literature we propose that some generalities about the structure of cognitive maps can be made. As this article is a synthesis of a large body of knowledge from different fields on cognitive mapping, there are many new technical terms. For the reader’s convenience, the first time a new term is used, it is explained and a reference to the relevant literature is given.

What is a cognitive map? A cognitive map can be described as a qualitative model of how a given system operates. The map is based on defined variables and the causal relationships between these variables. These variables can be physical quantities that can be measured, such as amount of precipitation or percent vegetation cover, or complex aggregate and abstract ideas, such as political forces or aesthetics. The person making the cognitive map decides what the important variables are which affect a system and then draws causal relationships among these variables indicating the relative strength of the relationships with a number between −1 and 1. The directions of the causal relationships are indicated with arrowheads. Cognitive maps are especially applicable and useful tools for modeling complex relationships among variables. With cognitive mapping the decision-makers’ maps can be examined, compared as to their similarities and differences, and discussed. In addition the effects of different policy options can easily be modeled. Maps can also be made with local people, who often have quite a detailed understanding of the ecosystem (Ozesmi, 1999a, Ozesmi, 1999b). Their input can be important for decision-making and for the public to accept the chosen solutions.

Cognitive maps are directed graphs, or digraphs, and thus they have their historical origins in graph theory, which started with Euler in 1736 (Biggs et al., 1976). In digraphs each link (line or connection) between variables (points or nodes) has a direction (Harary et al., 1965). Anthropologists have used signed digraphs to represent different social structures in human society (Hage and Harary, 1983). In ecology, Puccia (1983) used a signed digraph model for studying the relationships among benthic organisms. Axelrod (1976) was the first to use digraphs to show causal relationships among variables as defined and described by people, rather than by the researcher. He called these digraphs cognitive maps (term first used by Tolman, 1948). Many studies have used cognitive mapping to look at decision-making as well as to examine people’s perceptions of complex social systems (Axelrod, 1976, Bauer, 1975, Bougon et al., 1977, Brown, 1992, Carley and Palmquist, 1992, Cossette and Audet, 1992, Hart, 1977, Klein and Cooper, 1982, Malone, 1975, Montazemi and Conrath, 1986, Nakamura et al., 1982, Rappaport, 1979, Roberts, 1973).

Kosko (1986) modified Axelrod’s cognitive maps, which were binary, by applying fuzzy causal functions with real numbers in [−1, 1] to the connections, thus the term fuzzy cognitive map (FCM). Kosko was also the first to compute the outcome of a FCM, or the FCM inference, as well as to model the effect of different policy options using a neural network computational method (Kosko, 1987).

FCM has been used to model a variety of things in different fields: the physiology of appetite (Taber and Siegel, 1987), political developments (Taber, 1991), electrical circuits (Styblinski and Meyer, 1988), a virtual world of dolphins, shark, and fish (Dickerson and Kosko, 1994), organizational behavior and job satisfaction (Craiger et al., 1996), and economic/demographics of world nations (Schneider et al., 1998). Recent applications have included using expert knowledge to create FCMs that are combined with data mining of the world wide web (Hong and Han, 2002, Lee et al., 2002).

In ecology, the use of FCMs has been limited. Radomski and Goeman (1996) used FCM to suggest ways to improve decision-making in sport-fisheries management by sending questionnaires to experts asking them the important variables and the relationships between these variables. Although they emphasized the importance of knowledge concerning angler behavior when making management decisions, they did not incorporate the opinions of anglers in their model. Hobbs et al. (2002) used FCM to define management objectives for the Lake Erie ecosystem. Their FCM modeling process involved the participation of many experts and some members of the public, allowing for discussion and insight into the potential effects of different management actions. Ozesmi, 1999a, Ozesmi, 1999b first used FCM to analyze the perceptions about an ecosystem held by people in different stakeholder groups. Özesmi and Özesmi (2003) used FCM to analyze the perceptions of different stakeholder groups about a lake ecosystem in order to create a participatory management plan. Dadaser and Ozesmi, 2001, Dadaser and Ozesmi, 2002 used FCM to obtain the perceptions of different stakeholder groups in two wetland ecosystems in central Turkey. Recently applications of cognitive mapping have appeared in forest management. Skov and Svenning (2003) combined FCM with a GIS to use expert knowledge to predict plant habitat suitability in a forest. Hjortsø (2004) discussed the use of a cognitive mapping approach called strategic option development and analysis (SODA) to increase stakeholder participation in forest management in Denmark. Mendoza and Prabhu (2003) used cognitive mapping to examine the linkages and interactions between indicators obtained from a multi-criteria approach to sustainable forest management.

In this section we compare FCMs to other methods in ecological modeling and environmental management.

The use of expert systems is increasing in ecological modeling (i.e. Yamada et al., 2003). Expert systems require the construction of a knowledge base which is taken from the experts’ experience. Compared to most of these methods, FCMs are relatively quicker and easier to acquire from the knowledge sources, who do not usually think in equations. With FCMs you can have as many knowledge sources as wanted with diverse knowledge and different degrees of expertise. These knowledge sources can all be easily combined into one FCM. There is no restriction on the number of experts or on the number of concepts.

Structural equation modeling (SEM) (also known as causal modeling, covariance structural modeling, LISREL and others) is based on the statistical model developed by Jöreskog (1977). Causal relationships among the variables in the models are specified and tested with parameter estimation procedures, usually maximum likelihood. Structural equation modeling techniques are typically used to confirm or disprove an a priori hypothesized model. However, they can also be used as an exploratory modeling tool. Currently ecological applications of SEM are increasing (Iriondo et al., 2003, Shipley, 2000). Craiger et al. (1996) compared SEM with FCM. The limitations of SEM include nonconvergence of solutions and the inability to estimate parameters if the model and data are insufficient (under identification). In contrast FCMs are not concerned with parameter estimation but instead give qualitative information. Because of this FCMs facilitate pattern prediction, or changes in the behavior of the model. The person making the map decides on the strengths, these strengths can be changed easily and more simulations done to learn how the model changes with changing strengths of relationships. SEM often has the problem of model underidentification, especially with complex systems. In addition feedback loops must be removed from the model. However, FCMs can have unlimited complexity, including an unlimited number of concepts and reciprocal causal (feedback) relationships.

Multiattribute decision theory has been widely used in ranking a finite number of alternatives characterized by multiple, conflicting criteria or attributes (i.e. Luria and Aspinall, 2003). It allows for the measurement and aggregation of the performance of one or more options with respect to a variety of both qualitative and quantitative factors (criteria) into a single value. With multiattribute decision theory the alternatives need to be chosen and the factors and their weights. In contrast, with FCM the technique can be used to suggest the alternatives based on stakeholder input where each person making an FCM thinks of what is important and what should be included. In addition, FCM allows feedback loops.

Systems dynamics models use differential or difference equations to describe a system’s response to external factors (i.e. Håkanson and Boulion, 2003). They are used to model long term dynamic behavior of ecosystems. In contrast FCMs are not dynamic models. Systems dynamics models require a lot of empirical data about the ecosystem. FCMs are more appropriate for data poor situations. Although Stave (2003) used a systems dynamics model to engage stakeholder interest and build stakeholder understanding of the system and the basis for management decisions concerning water supply in Las Vegas, NV, USA, our multi-step FCM method is a participatory approach where stakeholders themselves are involved in building the model.

Interestingly studies of Uluabat Lake in Turkey have been done with both the multi-step FCM approach (Özesmi and Özesmi, 2003) and a systems dynamics model (Güneralp and Barlas, 2003). For the systems dynamics model, the authors admit that some of the data needed is either unreliable or not available and results need to be interpreted in that light. Although many simplifying assumptions are made, the systems dynamics model also includes social and economic components in addition to ecological, unlike most systems dynamics models. Based on the systems dynamics model the effect of different policy options was simulated. The authors concluded that the model is to be a laboratory where the effect of different policy options can be simulated and suggested ways in which the model could be improved. The systems dynamics model predicted that the lake ecosystem would not go to a turbid water state with few macrophytes. However, as of 2003, macrophytes have declined, especially submerged macrophytes. Although algae is not abundant, the water is turbid from suspended solids. The FCM model predicts that lake pollution continues to increase. The FCM was based on the perceptions of many stakeholders concerning the ecosystem. But the main difference in the modeling approaches is the purpose of the models. The FCM was used to develop a participatory management plan with the goals and objectives based on the stakeholders’ FCMs. Because the management plan was based on stakeholder input the stakeholders were able to take ownership of the plan and are working towards its goals.

Why choose FCM over other modeling methods? To answer this question, we must consider the issues of model complexity and the reason for the model. Obviously it is important to have a model that is complex enough for the problem to be solved; however data poor situations limit model complexity. Data is costly and often not available, especially in developing countries, where conservation efforts and management are important but not resolved. The multi-step FCM approach described herein is not obtained from empirical data but can be used for modeling perception and therefore social ideas of how systems work. This is essential for conserving an ecosystem where the support of many stakeholders is necessary. It is also useful for extension activities to educate stakeholders, if there are any misperceptions.

The main advantage of the multi-step FCM approach is that it is easy to build and gives qualitative results. It does not require expert knowledge in every field but can be constructed based on simple observations by anybody including indigenous or local people. It does not make quantitative predictions but rather shows what will happen to the system in simulations under given conditions of relationships. The model provides a better summary of relationships between variables instead of articulating how that relationship is in detail.

With FCMs the strengths and signs of relationships can be easily changed and simulations run easily and quickly. Thus they are ideal tools for theory development, hypothesis formation, and data evaluation. However, FCMs are not substitutes for statistical techniques; they do not provide real-value parameter estimations or inferential statistical tests.

Our multi-step FCM analysis approach includes the following steps:

  • (1)

    Drawing of cognitive maps.

  • (2)

    Determining if the sample size is adequate.

  • (3)

    Coding the cognitive maps into adjacency matrices.

  • (4)

    Augmenting individual cognitive maps and then adding them together to form stakeholder social cognitive maps.

  • (5)

    Analyzing the structure of individual and social cognitive maps using graph theoretical indices.

  • (6)

    Analyzing the differences and similarities in variables among stakeholder groups.

  • (7)

    Condensing complex cognitive maps into simpler maps for comparison purposes.

  • (8)

    Analyzing the outcomes of cognitive maps using neural network computation.

  • (9)

    Simulating different policy options through neural network computation.

This approach is described in detail in Section 2.

Section snippets

Obtaining cognitive maps

Cognitive maps can be obtained in four ways: (1) from questionnaires, (2) by extraction from written texts, (3) by drawing them from data that shows causal relationships, (4) through interviews with people who draw them directly. In this article, the first three methods are described only briefly because they have been covered in detail elsewhere.

Roberts (1976, pp. 333–342) told how to derive cognitive maps from questionnaires. He suggests using the opinions of many experts and also “lay”

Examples from our research

Ozesmi, 1999a, Ozesmi, 1999b used FCM to examine the views of different stakeholders in an environmental conflict. Government and non-governmental organization (NGO) officials wanted to create a national park to protect an important wetland delta ecosystem on the coast of north central Turkey while local people opposed this plan. In this case FCM was useful to obtain the opinions of the different stakeholders. The transparency of this method made it less suspect than questionnaires for local

Conclusions

In our experience the advantages of using multi-step FCM outweigh its disadvantages. In particular we have found it useful for obtaining the perceptions of different people in different stakeholder groups concerning an ecosystem. This allows for more appropriate conservation strategies and management plans to be made. We found that when FCMs were created with a standard methodology the structural indices of the maps were close to each other even across separate studies. This result suggests

References (77)

  • G.A. Mendoza et al.

    Qualitative multi-criteria approaches to assessing indicators of sustainable forest resource management

    For. Ecol. Manage.

    (2003)
  • S.L. Özesmi et al.

    An artificial neural network approach to spatial habitat modelling with interspecific interaction

    Ecol. Model.

    (1999)
  • S. Reimann

    On the design of artificial auto-associative neural networks

    Neural Networks

    (1998)
  • M. Schneider et al.

    Automatic construction of FCMs

    Fuzzy Sets Syst.

    (1998)
  • K.A. Stave

    A system dynamics model to facilitate public understanding of water management options in Las Vegas, Nevada

    J. Environ. Manage.

    (2003)
  • K. Yamada et al.

    Eliciting and integrating expert knowledge for wildlife habitat modelling

    Ecol. Model.

    (2003)
  • Axelrod, R., 1976. Structure of Decision: The Cognitive Maps of Political Elites. Princeton University Press,...
  • Bauer, V., 1975. Simulation, evaluation and conflict analysis in urban planning. In: Baldwin, M.M. (Ed.), Portraits of...
  • Biggs, N.L., Lloyd, E.K., Wilson, R.J., 1976. Graph Theory: 1736–1936. Clarendon Press,...
  • M. Bougon et al.

    Cognition in organizations: an analysis of the Utrecht Jazz Orchestra

    Admin. Sci. Quart.

    (1977)
  • S.M. Brown

    Cognitive mapping and repertory grids for qualitative survey research: some comparative observations

    J. Manage. Stud.

    (1992)
  • D.M. Buede et al.

    Convergence in problem solving: a prelude to quantitative analysis

    IEEE Trans. Syst., Man, Cybernet.

    (1993)
  • Carley, K., 1990. Content analysis. In: Asher, R.E. (Ed.), The Encyclopedia of Language and Linguistics, vol. 2....
  • K. Carley

    Extracting team mental models through textual analysis

    J. Org. Behav.

    (1997)
  • K. Carley et al.

    Extracting, representing, and analyzing mental models

    Social Forces

    (1992)
  • Colwell, R.K., 1997. Estimate S: Statistical estimation of species richness and shared species from samples. Version 5....
  • P. Cossette et al.

    Mapping of an idiosyncratic schema

    J. Manage. Stud.

    (1992)
  • J.P. Craiger et al.

    Simulating organizational behavior with fuzzy cognitive maps

    Int. J. Comput. Intel. Org.

    (1996)
  • Dadaser, F., Özesmi, U., 2001. Tuzla golu ekosistemi icin katilimci yonetim plani yontemi: bulanik biliseel haritalama...
  • Dadaser, F., Özesmi, U., 2002. Stakeholder analysis for Sultan Marshes ecosystem: a fuzzy cognitive approach for...
  • Dadaser, F., 2002. Katilimci Yonetim Yaklasimiyla Tuzla Golu Ekosistem icin Cevre Yonetim Plani (An environmental...
  • J.A. Dickerson et al.

    Virtual worlds as fuzzy cognitive maps

    Presence

    (1994)
  • C. Eden et al.

    The analysis of cause maps

    J. Manage. Stud.

    (1992)
  • Eden, C., Jones, S., Sims, D., 1979. Thinking in Organizations. The Macmillan Press,...
  • Gower, J.C., 1985. Measures of similarity, dissimilarity, and distance. In: Kotz, S., Johnson, N.L., Encyclopedia of...
  • Hage, P., Harary, F., 1983. Structural Models in Anthropology. Oxford University Press, New...
  • Harary, F., Norman, R.Z., Cartwright, D., 1965. Structural Models: An Introduction to the Theory of Directed Graphs....
  • J.A. Hart

    Cognitive maps of three Latin American policy makers

    World Politics

    (1977)
  • Cited by (714)

    View all citing articles on Scopus
    View full text