Blog
7/11/2023
Understanding how a group of people perceives a problem is of multiple interest. In the context of finding solutions to complex or ambiguous situations, how can we ensure that the relevant actors develop a common understanding of the issues they are dealing with? In a world where the complexity of issues is growing, and the need for interdisciplinary approaches and cross-sectoral action is widely recognized, understanding multiple perspectives and coming together to build a common vision is essential. Whether it’s a question of better understanding the social stereotypes that hinder disease prevention or people’s access to healthcare, or synthesising different viewpoints on urban development and the loss of natural environments, there’s no shortage of issues and multiple points of view. Concept Mapping (or Group Concept Mapping) is one method that can help address such challenges.
The concept mapping method to which we refer here is the one developed by Trochim (1989) in the field of program evaluation and planning. The aim of this method is to get a group of individuals to visualize their collective understanding of a particular issue. It is a mixed method combining brainstorming and free sorting of ideas with multivariate statistical analysis to map participants’ representation of a problem. Ultimately, it generates various visualizations, notably the concept map, which illustrates the issues identified and brings out structuring themes, making it possible to synthesize the common understanding and envisage the development of solutions1.
“The aim of this method is to get a group of individuals to visualize their collective understanding of a particular issue.”
For the organizers of a concept mapping exercise, the aim is to obtain a visual representation of how a group of people view a given issue. The first step is to define a target question to gather the material of interest (e.g., what the stereotypes associated with depression are). In the second step, respondents are asked to provide individual answers to the question (e.g., depression is uniquely determined by life circumstances). Once this collection of ideas is complete, they are first cleaned – including orthographic correction, separation of sentences containing two ideas, and deletion of invalid ideas – to then form a representative subsample of the initial ideas, generally comprising around 40 to 50 items.
The third step includes a free sorting task and a rating task. Each participant is asked to group together ideas he or she considers conceptually similar (e.g., antidepressants change personality & antidepressants must be taken for life), and then to rate each idea according to previously established criteria (e.g., the “harmful” and “widespread” characteristics of each stereotype). The data generated in the sorting stage is represented in the form of a similarity matrix of size n by n, where n represents the total number of sorted ideas. Each cell in the similarity matrix indicates the number of times two ideas were grouped together across the participants.
The fourth step consists of generating the concept map by combining two statistical methods: a cluster analysis and a dimensionality reduction. The first allows us to delimit clusters of ideas that are conceptually close according to a significant number of respondents - the second leads to a projection of the ideas into a two-dimensional space. The projection of those clusters in the 2-dimensional space is what is called the concept map. The number of clusters in the final map is determined by quality indices (Halkidis et al., 2001) and the analyst’s theoretical understanding of the problem2.
The final stage involves the interpretation of the map. Participants are asked to name the clusters in an inductive process, and if necessary to draw structuring axes to facilitate interpretation. The clusters are considered to be general concepts, constituting the group's collective representation of the problem (see fig. 1). Concept mapping thus differs from other forms of analysis, where it is usually the people in charge of the research who themselves determine the names of the clusters and the axes.
Alongside the concept map, participants can also rate each idea according to previously established criteria (e.g., on a scale of 1 to 10, how harmful and widespread stereotypes are). Based on these ratings, the ideas are projected onto a scatter diagram, called a GoZone plot, whose axes represent the criteria. This makes it possible to qualify not only the ideas, but also the clusters, according to the selected criteria (see fig. 2).
This mixed-method approach is often used in participatory research, program evaluation and planning. At Polygon, we have enhanced this method, which we describe in this article. The various projects which used our solution will provide an overview of the possibilities offered by CM*.
Notes:
1. This technique differs from the one developed by Novak in the 1970s and popularized in the field of educational science (Novak, 1977). In this second method, the concept map is a diagram representing the relationships between different concepts. It is to the Novak method that we most commonly refer when we speak of Concept mapping. Trochim’s method is also sometimes known as Group Concept Mapping, to distinguish it from Novak's method. However, the term Concept Mapping, rather than Group Concept Mapping, is generally used in scientific articles using Trochim’s method
2. In the method developed by Trochim, items are first projected into a two-dimensional space by non-metric multidimensional scaling (nMDS), and the clustering analysis is then performed on the coordinates obtained by nMDS. This approach is, however, questioned by Péladeau et al, (2017), and at Polygon, we use a different approach discussed in those following articles: An alternative approach to the concept mapping method: clustering analysis on original distances & Dimensionality reduction: Alternative approaches to concept mapping.
Bibliography:
Halkidi, M., Batistakis, Y., & Vazirgiannis, M. (2001). On clustering validation techniques. Journal of Intelligent Information Systems, 17 (2–3), 107–145. https://doi.org/10.1023/A:1012801612483
Péladeau, N., Dagenais, C., & Ridde, V. (2017). Concept mapping internal validity: A case of misconceived mapping? Evaluation and Program Planning, 62, 56–63. https://doi.org/10.1016/j.evalprogplan.2017.02.005
Trochim, W. M. (1989). An introduction to concept mapping for planning and evaluation. Evaluation and Program Planning, 12 (1), 1–16. https://doi.org/10.1016/0149-7189(89)90016-5
Tools & support solutions
Participatory research
Research methods
Collective intelligence
Focus group
Delphi
