iModeler manual: a quick guide for fuzzy cognitive modelling

Submitted by Anneli Sundin | published 27th Jun 2015 | last updated 16th Nov 2015
capi 2 0 - climate adaptation.

Vista Aerea de una zona del Bosque Modelo Chiquitano, Bolivia

Summary

Cognitive maps are qualitative graphical models of a system, consisting of a set variables linked together by causal relationships. A cognitive map can be made of almost any system or problem, and therefore particularly useful for creating models based on people’s knowledge (experts and lay people). As a tool to express people’s mental models, it is useful as a brainstorming tool, an alternative to other “discussion support tools” such as the mind map (associative relationship among ideas). A form of cognitive maps, the concept map2 (Novak, 1990), typically allow to build interconnected build propositions, i.e. concept-relation-concept (e.g. “farmer manages its farm”) from people’s discourses, and as such can be built quickly without any technical skills, and is readily understandable by a technical or non-technical audience. Cognitive maps are important in decision- making since 1) they help raise self-awareness of complexities 2) help communication by making tacit knowledge explicit 3) they reveal how individual perceptions of reality shape choices 4) they make explicit critical choices parameters and trade-offs, encouraging negotiation and promoting win-win solutions.

Concept maps, like influence diagrams or mind maps, are essentially semantic or graphic, therefore they are not meant for numerical processing e.g. for scenario analysis, for assessing the sensitivity of parameters, or for impact assessment. Causal mapping like influence diagrams or binary cognitive maps (or bayes nets which allows to specify joint probabilities) are a step further in a more quantitative (yet fuzzy) assessment of complex systems. Kosko (1986) modified binary cognitive maps (Axelrod, 1976) by applying fuzzy causal functions, i.e. assigning real numbers in [−1, 1] to the relationship between factors, thus the term fuzzy cognitive map (FCM) (Özesmi and Özezmi, 2004). Kosko (1987) 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 neural networks. Modern FCM combine aspects of fuzzy logic, neural networks, semantic networks, expert systems, and nonlinear dynamical systems. FCMs are increasingly used to address a great diversity of problems, from knowledge representation, fuzzy control, engineering, approximate reasoning, strategic planning, management medical decision making, game theory, and data mining analysis. Closer to our research interests, FCM have been used for environmental policy (Kontogianni et al, 2012), developing freshwater future scenarios in Europe (VanVliet et al, 2010), analyzing deforestation in the Brazilian amazon (Kok, 2009), Ecosystem conservation and ecological modeling (Ozesmi and Ozesmi, 2003; 2004).

Given the context of our intervention (scarcity of data, multi-stakeholder decision-making, science- society processes) we judged that FCM was promising for developing, jointly with stakeholders, scenarios and alternatives and evaluate their consequences. However going from a positivist paradigm (e.g. quantitative models) to a constructivist one (e.g. qualitative or semi quantitative models) is not obvious, as those of us trained with a positivist mindset may have to be convinced of the benefits of a more “fuzzy” approach to modeling. Therefore, the approach is similar to grounded theory or qualitative social research, where “reality” is context-dependent scenarios of possible developments are uncertain.

This section is based on our efforts in implementing FCM in EcoAdapt sites, to provide methods, best practices and tips, highlighting their strengths and weaknesses. It has benefited from feedback from CSO staff during on-site training in FCM modelling by CIRAD in Argentina and Bolivia in October and November 2013.  We used the user-friendly iModeler software (http://www.consideo.com/) which implements non-standard FCMs, so we provide guidelines specific to this software. We also present the FCMs obtained from the socio-ecological PARDI conceptual models for EcoAdapt Argentina, Bolivia and Chile sites.

Suggested citation

Leclerc, G. 2014. iModeler manual: a quick guide for fuzzy cognitive modelling. EcoAdapt Working Paper Series N°2 Adaptation to climate change for local development.