By Pete Barbrook-Johnson and Alexandra S. Penn
What are some effective approaches for developing causal maps of systems in participatory ways? How do different approaches relate to each other and what are the ways in which systems maps can be useful?
Here we focus on seven system mapping methods, described briefly in alphabetical order.
1. Bayesian Belief Networks: a network of variables representing their conditional dependencies (ie., the likelihood of the variable taking different states depending on the states of the variables that influence them). The networks follow a strict acyclic structure (ie., no feedbacks), and nodes tend to be restricted to maximum two incoming arrows. These maps are analysed using the conditional probabilities to compute the potential impact of changes to certain variables, or the influence of certain variables given an observed outcome.
2. Causal Loop Diagrams: networks of variables and causal influences, which normally focus on feedback loops of different lengths and are built around a ‘core system engine’. Maps vary in their complexity and size and are not typically exposed to any formal analysis, but are often the first stage in a system dynamics model.
3. Fuzzy Cognitive Mapping: networks of factors and their causal connections. They are especially suited to participatory contexts, and often multiple versions are created to capture diverse mental models of a system. Described as ‘semi-quantitative’, factors and connections are usually given values, and the impacts of changes in a factor value on the rest of the map are computed in different ways.
4. Participatory Systems Mapping: a network of factors and their causal connections, annotated with salient information from stakeholders (eg., what is important, what might change). Maps tend to be large and complex. They are analysed using network analysis and information from stakeholders to extract noteworthy submaps and narratives.
5. Rich Pictures: a free-form drawing approach in which participants are asked to draw the situation or system under consideration as they wish, with no or only a handful of gentle prompts. This method is part of the wider group of soft systems methodologies.
6. System Dynamics: a network of stocks (numeric values for key variables) and flows (changes in a stock usually represented by a differential equation), and the factors that influence these. Normally, these maps are fully specified quantitatively and used to simulate future dynamics.
7. Theory of Change Maps: networks of concepts usually following a flow from inputs, activities, outputs, and outcomes to final impacts. Maps vary in their complexity and how narrowly they focus on one intervention and its logic, but they are always built around some intervention or action. Maps are often annotated and focused on unearthing assumptions in the impact of interventions.
How do these methods relate to each other?
The following three figures show how these methods relate to each other. While individual projects could use any of these methods in a different way, these figures give a rough sense of where these methods sit in relation to one another, and what some of the most important axes on which to differentiate them are.
The first figure looks at the overall focus and nature of the different system mapping methods.
The second figure focuses on the mode and ease of use of the different system mapping methods.
The third figure presents the outputs and analysis the different system mapping methods produce.
How can systems mapping be useful?
We next suggest five broad types of use, which also apply to most types of modelling or analysis.
1. Helping us think: system maps of all types force us to be more specific about our assumptions, beliefs, and understanding of a system. Many types of systems mapping also force us to structure our ideas using some set of rules or symbols (ie., creating boxes and lines to represent concepts and their relationships). This will introduce simplifications and abstractions, but it will also make explicit our mental models.
2. Helping us orient ourselves: a systems mapping process will often also help us orient ourselves to a system or issue. Whether a map helps us see our, and others’, positions in the system, or whether it helps us quickly develop a fuller understanding of an issue, we will be better oriented to it. This helps people navigate the system better, be aware of what else to think about when considering one part of a map, or know who is affected and so should be included in discussions.
3. Helping us synthesise and connect information: the more flexible types of mapping are particularly good at bringing together different types of data, evidence, and information. They can all be used to inform the development of a map, making connections that would not otherwise be possible. Different types of visualisation, hyperlinking, and map structure can also be used to help people return to the information underlying a map.
4. Helping us communicate: whether we build maps in groups, or alone, and then share them, all system maps should help us communicate our mental models and representations of systems. The process of mapping with others, and the discussions it generates, unearths a multitude of assumptions which can then also be challenged and unpicked. The end product of a mapping process can also help us communicate our ideas about a system. Maps can become repositories for our knowledge which can be accessed by others, and updated, becoming a living document.
5. Helping us extrapolate from assumptions to implications: systems mapping approaches which can be turned into simulations, or which can be analysed in a formal way, also allow us to follow through from the assumptions we have embedded in them, to their implications.
Are there other methods that you use to develop causal maps of systems and that can be used in participatory ways? What’s the main value that you have found in systems mapping? Do you have other lessons to share from your experience of systems mapping?
To find out more:
Barbrook-Johnson, P. and Penn, A. S. (2022). Systems Mapping: How to build and use causal models of systems. Palgrave-Macmillan: Cham, Switzerland. (Online – open access): https://link.springer.com/book/10.1007/978-3-031-01919-7
Biography: Pete Barbrook-Johnson PhD is a social scientist and complexity scientist working on a range of environmental and energy policy topics, using systems mapping, agent-based modelling, and other related approaches. He is a Departmental Research Lecturer at the University of Oxford in the UK and a member of the Centre for the Evaluation of Complexity Across the Nexus (CECAN) hosted by the University of Surrey in Guildford, UK.
Biography: Alexandra S. Penn DPhil is a complexity scientist working on combining participatory methodologies and mathematical models to create tools for stakeholders to understand and ‘steer’ their complex human ecosystems. She is a Senior Research Fellow at the University of Surrey and a member of the Centre for the Evaluation of Complexity Across the Nexus (CECAN) hosted by the University of Surrey in Guildford, UK.
16 thoughts on “Seven methods for mapping systems”
1. One other thing worth mentioning is that Pete and Alexandra’s excellent book is available in (free) downloadable pdf form here; https://link.springer.com/book/10.1007/978-3-031-01919-7#toc Well done for making the book so accessible!
2. Their book has a chapter titled “What Data and Evidence Can You Build System Maps From?” and a section therein titled “Using Qualitative Data to Build Your Map” Up until now this has involved quite a lot of time consuming manual coding of text material, albeit helped by the different software packages mentioned in this section. But with the advent of ChatGP and the like, extraction of causal relationships from text is now much simpler and quicker, though not without its pitfalls. Steve Powell, mentioned in this same section,has written about this new capacity here https://www.causalmap.app/post/chatgpt-is-changing-how-we-do-evaluation-the-view-from-causal-map and I have written more generally about ChatGPT’s capacity for qual analysis here: https://mande.co.uk/2023/lists/software-lists/using-chatgpt-as-a-tool-for-the-analysis-of-text-data/
fyi, rick davies
Thanks Rick – I would be very interested in exploring more what Steve has been thinking about. I agree the potential to quickly build robust causal system maps with tools like ChatGPT is very exciting. I see this as another way of doing ‘data-driven’ systems mapping, like we do here, but with time series quant data – https://www.inet.ox.ac.uk/publications/no-2022-26-using-data-driven-systems-mapping-to-contextualise-complexity-economics-insights/
A really helpful introduction, which is expanded on clearly and engagingly in the full book. We’ve learned so much from Alex and Pete’s work in the development of new guidance for systems mapping in population health research, policy and practice – coming very soon!
Thanks Ben! I think your more bespoke / discipline focussed guidance is invaluable – need this in other areas too.
I was just talking about some of the chapters in this book with a colleague and mentioned it as a great primer and introduction. Clearly written and informative for those surveying the field! Congrats Alexandra and Pete and thanks for your contribution.
Thanks Steven – means a lot from someone who has done so much great work in this area!
Interesting summation of a competency I think should be in every leaders toolkit no matter what their field for the reasons you state. Two other systems mapping approaches I use are called Conversation Mapping (a version of rich picturing for those to shy or unable to draw) that organises the perceptions multiple stakeholder have of a systems (as in SSM – soft systems methodology) and Coherence Mapping which explores the relationships between nodes in a system using the question, ‘what do you need from me to achieve your contribution to the whole’s fulfilling its purpose?’ The Coherence map works well in an ‘is-ought’ exploration or when planning a new intervention. Both approaches reveal many of the hidden assumptions that stakeholders hold about the systems they are involved in. The answers you get to your concluding questions would make a great follow up post.
Thanks – would love to know more about these two – do you have any favourite resources / writing on them?
They are both part of the courses we teach and I’d be happy to send you the appropriate course notes if you provide me with a contact address. My email is firstname.lastname@example.org There is a YouTube video of me introducing Conversation Mapping about 12 years ago to a group in California at https://youtu.be/uqwL4k2easU
Thanks – will email
A timely post for me! Thanks for sharing! Have you used giga-maps much in your approach and if so, how does the method work for you alongside the other maps you mention? I haven’t used the method myself so don’t know too much about it. But I’m interested in how to qualitatively map the complexity of a system. Rich pictures I find is a really great and simple way to do this. It resonated with me. Thanks.
Hi, I have not used gigamaps before, and we dont focus on them in the book, but we do point to this resource, which is a good place to start I think – https://systemsorienteddesign.net/what-is-gigamapping/.
Nice approach! The iMODELER as a tool combines system dynamics, CLDs (causal loop diagrams), FCM (fuzzy cognitive maps) and to an extent the sensitivity model by Prof. Vester: https://www.consideo.com/files/consideo/pdfs/papers/eng/Why_iMODELER.pdf
Nice, I had not come across this before, will check it out. What do you think the USP (unique selling point) of this software is, compared to others?
The paper should list the software’s unique selling points. It all started from an EU research project to make system dynamics easier. We than continued to explore qualitative modeling approaches and developed our own which started with a rough weighting of connections using the attributes “weak”, “medium”, and “strong” and then continued to go beyond the Fuzzy Cognitive Maps …. It offers a bionic view switching perspectives to handle very large models and to work collaboratively with teams on the same model. The iMODELER also features process and resource factors and some algorithm to identify constraints according to the Theory of Constraints and to come up with optimal sets of parameters for a given goal. And it offers the use of an expert system (know-why.net) to get some proposals for potential factors which as a matter of fact we are currently reviewing in the light of ChatGPT and similar tools. So I am really excited to read the comment from Rick Davies. But that is just the tool – you emphasize the experiences from using the different approaches. We use 4 questions to facilitate the collection of crucial factors translating the natural language from a workshop into cause and effect relations. After a third of the time collecting the arguments we use another third to weight the connections and then the final third to look at the system, its loops and the insight matrices that tell what factors (targets, measures, obstacles, etc.) seem to be the most effective short, medium, and long term. If there is time for more and one wants to know how and with what likelihood things are probably going to develop the de facto causal loop diagram can be translated into a quantitative system dynamics program without the need to switch to a stock and flow diagram. However, the development of a system dynamics model collecting data and developing formula is usually not a participative effort.
Thanks – I will take a close look, for one!