CMSE 495
This is the webpage for CMSE495 Data Science Capstone Course (Spring 2022)
In-Class Exercise: Team Fuzzy Cognitive Mapping
Fuzzy Cognitive Map representation of a coupled social and ecological system
In class today we are going apply some methods from participatory systems modeling to understand the drivers and dynamics of project success. We will begin with an introduction to systems thinking through a participatory complex systems modeling approach known as Fuzzy Cognitive Mapping (FCM). As a group, we will then define a set of intereacting model components affecting, and affected by, project success. Next, we will break into teams to build team-level models. We will use these models to run scenarios with the goal of building intuition and facilitating communication and learning among team members. We will then report findings back to the class and discuss.
Everyone will be given 2 class periods (today and next Wednesday) to finish this project, and teams will present their work at the end of the second class.
Agenda (2 Days = 160 Minutes)
- (10 minutes) Introduction to Systems Thinking and FCM
- (20 minutes) Participatory Generation of Model Components
- (?? minutes) Team Modeling Exercise
- (5 minutes) Daily Team Check-in
0. “All Models Are Wrong…”
“…But some are useful,” so the saying goes (Box, 1976). At the beginning of the semester, we told you each project would look like this:
This, of course, is wrong. The truth is far messier. Today we will attempt to move towards it. To do this, we need to trade our Newtonian, linear thinking for that of Donella Meadows, a nonlinear, systems-orientated approach (Meadows, 2008).
References
Box, G.E.P. 1976. Science and statistics. Journal of the American Statistical Association. 71(356):791-799.
Meadows, D. 2008. Thinking in Systems: A Primer. Chelsea Green Publishing. While River Junction, VT.
1. Introduction to Mental Modeler
Mental Modeler is a software that will allow us to build FCM representations of our projects, specifically those elements that dynamically contribute to its success as so defined with respect to each project, the community partner and its representatives, the team members, and associated stakeholders. The following is a general introduction to the ideas informing the software.
Fuzzy Cognitive Maps
Fuzzy Cognitive Maps (FCMs) combine aspects of dynamical systems, graph neural networks, and fuzzy logic to model and simulate complex systems. An FCM is a subjective representation of internal knowledge and its structure (Jones et al., 2011). Such models encode a deterministic structure that is believed to emulate an approximation of human thought, and thereby allow simulation of the human process of causal inference (Jones et al., 2011). Unlike classical causal inference that relies on DAGs (directed acyclic graphs) and strict statistical assumptions, FCM allows for more complex conceptions of causality via weighted and directed graphs whose topological structure may include cycles, thereby encoding feedback mechanisms. The first participatory application of FCM investigated decision-making processes of army officers engaged in complex war games (Klein & Cooper, 1982). With respect to the wicked problems paradigm of Rittel and Weber (1973) that we met early in the semester, mental modeling excels in its ability to represent diverse sets of knowledge (Biggs et al., 2011) that may be aggregated into a collective cognitive map (Gray, 2016). The method of forming a collective map is believed to increase stakeholder knowledge of a system, to enable critical reflection on their conceptions, and to facilitate social learning (Papageorgiou, Markinos, & Gemptos, 2009; Gray, 2016).
References
Biggs, D., N. Abel, A. Knight, A. Leitch, A. Lanston, and N. Ban. 2011. The implementation of crisis in conservation planning: Could ‘mental modeling’ help? Conservation Letters. 4(3):169-183.
Jones, N., H. Ross, T. Lynam, P. Perez, and A. Leitch. 2011. Mental models: An interdisciplinary synthesis of theory and methods. Ecology and Society. 16(1).
Klein, J.H., and D.F. Cooper. 1982. Cognitive maps of decision-makers in a complex game. Journal of Operational Research Society. 33(1):63-71.
Gray, S. 2016. Private correspondence.
Gray, S., E. Zanre, and S. Gray. 2014. Fuzzy cognitive maps as representations of mental models and group beliefs. From Fundamentals to Extensions and Learning Algorithms. Springer.
Papageorgiou, E., A. Markinos, and T. Gemptos. 2009. Application of fuzzy cognitive maps for cotton yield management in precision farming. Expert Systems with Applications. 36(10):12399-12413.
Rittel, H. & M. Weber. 1973. Dilemmas in a General Theory of Planning. Policy Sciences. 4:155-169.
2. Whole-Class Component Generation
Brainstorming
✅ DO THIS: Take 3 minutes and write a list of concepts that affect project success. These components need to be able to increase and decrease. Think of them as stocks, like fish in a lake: Fish populations can shrink and grow. Project success (a fuzzy concept), by analogy, can improve or decline. Use the nodes of this example FCM diagram for inspiration:
We note for later that blue arcs in this visualization represent positive correlations and the orange represent negative.
Share with Whole Class
✅ DO THIS: Share ideas as a class. We will discuss and refine these concepts together.
3. Team Modeling Exercise
Exercise
The goal of this exercise is to facilitate social learning and communication about current understanding of your team projects (an instance of single-loop learning). Next, you will have the opportunity to critically reflect on these beliefs and assumptions (an instance of double-loop learning). Through the process, you will identify the drivers and dynamics of project success through a participatory modeling exercise. The exercise consists of:
- Producing team-level models constructed in a participatory manner (Days 1 and 2)
- Simulating and discussing a subset of possible scenarios (Days 1 and 2)
- Sharing insights with the class (Day 2)
Team Roles
Teams will enter breakout rooms. When your team arrives, you will designate roles defined as follows. Please note that everyone will participate in the activity, and that these roles may be shared among members if necessary.
-
Facilitator: The facilitator guides the model generating process. They will:
a. Ask the team which components have a causal effect on which; and,
b. For each causal relation $X \to Y$, they will ask:
- “When component $X$ goes up, does component $Y$ go up or down?”
- “By how much: A little, some, or a lot?”
-
Modeler: The modeler’s job is to translate the results of the discussion into Mental Modeler. This includes:
a. Navigating to the Mental Modeler website and logging in using the provided credentials; and,
b. Adding concepts to the cognitive map using the Mental Modeler GUI; and,
c. Assigning directional relations and weights using the GUI sliders; and
d. Saving a screenshot of the diagram, downloading and saving the
.mmpfile, and exporting model structure to CSV.
Everyone else will be known as participants.
Note: The .mmp file has all of your modeling progress and will allow the team to pickup next week where you left off today. We will check-in as a class for the last 5 minutes to assess progress and challenges, as well as to brielfly summarize interesting results thus far. At the end of the second day, you will submit your CSV to the instructors.
✅ DO THIS: In your team breakout room, use the components we arrived at as a group to generate an FCM addressing the dynamics affecting project success.
✅ DO THIS: Explore scenarios and record your observations.
✅ DO THIS: Coordinate a 5 minute (max) presentation, and be ready to share what your entire team did. We will have presentations at the end of next Wednesday.
✅ DO THIS: Download and submit CSV of your team FCM.
Written by Nathan Brugnone, Michigan State University
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.