Why complex problems need abductive reasoning

By Mariana Zafeirakopoulos

Mariana Zafeirakopoulos (biography)

How does the way we approach complex problems differ from how we approach problems that are familiar or obvious?

In this i2Insights contribution, I explore four kinds of reasoning:

  • Deduction
  • Induction
  • Abduction
  • Design abduction.

Design abduction is the brain-child of Professor Kees Dorst (2015). In simplified terms, these different kinds of reasoning can be compared as follows (Watson and Dorst, 2022, p. 3; taken from Dorst, 2015, pp. 46-49):

Deduction – solid reasoning from cause to effect…
WHAT + HOW leads to ???

Induction – discovering patterns…
WHAT + ??? leads to OUTCOME

Normal abduction – solid problem solving based on experience
??? + HOW leads to OUTCOME

Design abduction – two unknowns lead to a process of creative exploration
??? + ??? leads to OUTCOME 

For familiar or obvious problems, we apply best practice – tested approaches that have been proven to work in similar problem situations. Deductive and inductive reasoning respectively deal with known facts and established definitions, or patterns that emerge over time.

Deductive reasoning applies to cause and effect, often in the form of ‘if’ and ‘then’ statements. As in ‘if’ something is true ‘then’ something else must follow or be true. For example, if a sports team beats their competitor, then they will move to the next round of the competition.

Inductive reasoning is different from deduction in that it combines observations with information to reach a conclusion. It involves generating conclusions from the specific to the general. For example, I observe a baby speak its first words at 12 months, I see other babies do this at around the same age and conclude that all babies speak their first words by the time they are 12 months old.

For complex problems, we cannot rely on best or good practice. Trial and error does not work for complex problems because the nature of the problems is different. Complex problems are open, networked and dynamic.

They are open in that they are often boundaryless, or the boundaries we draw change the nature of the problem. They are networked in that they are connected to other problems or systems, whereby a change in one area affects another system. They are dynamic in that they are often changing or mutating in incremental or often imperceivable ways until the change can be recognised.

Complex problems are the realm of emergent practice. Often, the problem – or how we could have responded – only makes sense in hindsight. Working with these emergent, unknown and unknowable spaces means that we draw from other ways of knowing that are not embedded in facts. When working with emergent practices, we often draw from intuition and lived experience but also harness creativity and make mental leaps to explore possibilities, ideas and the unreal. These tools are a legitimate form of rationality called abductive reasoning.

Abductive reasoning was originally defined by the Pragmatist philosopher, Charles Sanders Peirce (pronounced Purse) as the logic or method to scientific discovery. Peirce realised that induction and deduction were not the forms of reasoning that helped scientists discover new things, or to create theories or hypotheses. Rather, induction and deduction were forms of reasoning that supported abduction.

The definition of abductive reasoning has evolved since Peirce’s original conceptualisation. One example of how abduction has evolved for complex contexts is Dorst’s work on ‘design abduction’. Design abduction is a process of creative exploration that involves establishing the outcome of a problem first, and then coevolving the ‘how’ we get to that destination with ‘what’ we need to get there.

For example, we might like to reimagine the relationship between police and community in ways that are less punitive or oppositional in favour of more inclusive and rewarding ways of promoting care, safety and security. To achieve this desirable outcome, we might need changes in policy or legislation, and dialogue sessions with community and policing as equal stakeholders to imagine a different relationship. The ‘what’ and ‘how’ work together forming probes into the problem space. These observed actions might lead to exploring new and different means and to achieve the desired outcome.

Others have explored the evolving nature of abduction in terms such as:

  • the logic of what might be (Kolko 2010),
  • the argument to the best explanation (Martin in Kolko 2010),
  • creative leap (Veen 2021).

What design abduction and these other definitions have in common for working on complex problems, includes:

  • Practicing sensemaking. This is a motivated and continued effort to understand connections. Sensemaking is the opposite of reductionism (which is used in analysis of obvious problems).
  • Legitimising constructivism. Constructivism, in contrast to positivism, legitimises human and personal experiences as ways of creating knowledge and meaning through human interactions.
  • Engaging with creative problem solving. This is a legitimate form of rationality needed alongside other forms of reasoning. Creative problem solving helps co-evolve our understanding of complex problems and interventions to those problems through processes of inquiry.
  • Adopting knowledge co-production. This is an inclusive way for researchers to bring in diverse stakeholder voices to reflect on the outcomes they want from the problems they face as a collective.
  • Shaping Futures. Abductive reasoning can help us form futures, not just inform them. Forming futures needs new ways of framing problems and new frames for seeing possibility, unlocking us from how we currently see, experience and engage with problems.

Abductive reasoning is needed in addressing complex problems because operating in the realm of the unknown, unknowable or unimaginable requires more than fact, definition and pattern-recognition. Although these are critical functions of reasoning, it is also important to acknowledge human experience, creative exploration and intuition as appropriate forms of rationality for dealing with complex problems.

Do these features of abductive reasoning in complex contexts resonate for you? Are there other features that you think should be included? How does the idea of considering human experience and intuition as a rationality feel for you? I would love to read your thoughts.


Dorst, K. (2015). Frame Innovation; Create New Thinking By Design. MIT Press: Cambridge, Massachusetts, United States of America.

Kolko, J. (2010). Abductive Thinking and Sensemaking: The Drivers of Design Synthesis. MIT Design Issues, 26, 1.

Veen, M. (2021). Creative Leaps in Theory: The Might of Abduction. Advances in Health Sciences Education, 26: 1173-1183.

Watson, R. and Dorst, K. (2022). Pragmatism, Design and Public Sector Innovation: Reflections on Action. In, Lockton, D., Lenzi, S., Hekkert, P., Oak, A., Sádaba, J., Lloyd, P. (eds.), Design Research Society Conference, DRS2022, 25 June – 3 July: Bilbao, Spain. (Online – open access): https://doi.org/10.21606/drs.2022.778

Biography: Mariana Zafeirakopoulos MA is a PhD Candidate at the TD (Transdisciplinarity) School, University of Technology Sydney (UTS), Australia, where she is researching the role of design in enabling knowledge coproduction (specifically, knowledge integration) to create preferred futures in national security contexts. She is affiliated with the University of Sydney, Australia, as an Academic Coordinator in Design; Charles Sturt University, Australia, as an Adjunct Lecturer on their Strategic Intelligence program; and as a casual academic with the TD School at UTS.

27 thoughts on “Why complex problems need abductive reasoning”

    • Hi Bethany

      Thank you for your energy and interest! I was percolating on your question and I’d love to continue the conversation with you. Where my current thinking is at, is that sensemaking is about endorsing what we see ‘between the lines’ – our lived experience, our intuition. It is perhaps a different way of allowing us to name our biases in a way that can bring them out into the open and allow for engagement with others about the meaning we make from the situation at hand.

      If analysis is the act of taking components apart and studying them in detail in its various form (e.g. how a car works), sensemaking then is more about how we interpret the form that is in front of us and the value it brings (e.g. why have a car, the values you bring when you are choosing a car ie greenness v speed or aesthetic v comfort) – what meaning do we collectively attribute to this. I don’t know if the car metaphor works very well, but what I think it reveals is the importance that reflexivity has in sensemaking as an ongoing process, checking the why against the what and the how.

      Keen to hear about what yours and others understanding of sensemaking is in emerging complex problems!

  1. Dear Mariana,

    thank you for an interesting post and discussion. I am playing around with vocabulary these days. What you describe as complex problems seems to be the same or similar to what many people call ‘wicked’ problems. I wonder if you came across this distinction and how do you relate your work to ‘wicked’ challenges? thank you.

    • Dear Varvara,

      What a great question! I have come across Rittel and Webber’s work in wicked problems. I think there’s a lot of cross over and commonality. In particular, the idea that you cannot solve complex / wicked problems but only intervene in them and shift and evolve them, I think are synonymous between the two. I suspect this has to do with underpinning systems theory influence, among other theories.

      The main reason why I prefer the language of complex problems over wicked problems is for 2 reasons: 1. the Cynefin framework. I like how the Cynefin framework distinguishes between different types of problems. For my ‘home’ discipline (that of Intelligence Analysis) having a framework like Cynefin is really helpful because it provides a way of stepping into the problem space, a way of seeing problems, offering different approaches according to the type of problem. I found this dynamism between obvious, complicated, complex, chaotic problems really useful because many practices assume an ‘obvious’ problem orientation looking to best practice, when actually, the problem they might be stepping into might be complicated (realm of good practice) or complex (the realm of emergent practice). These comparisons are quite interesting for my PhD research into how Strategic Intelligence analysis practices might need to be evolved to grapple with this kind of future emergence, because much of what Strategic Intelligence does is operate within a problem orientation of best practice. So the problem framing is at odds.

      The second reason is Professor Kees Dorst’s work on design and problem framing. Given my interest in the role of design practice, I wanted to lean on theories and approaches that looked at complexity from that disciplinary perspective. Dorst defines complex problems as those that are networked, dynamic and open. For me, this way of thinking made sense to me in connection with the Cynefin framework.

      I feel that whether you call a complex problem wicked or complex, there are fundamental commonalities. I wonder whether which definition one chooses depends on the purpose of how one discusses the nature of the problem. Ie organisational change v. leadership v. an analyst in the system working out how to ‘intentionally’ figure out what type of problem (or which dimension) they are seeking to inform.

      I’ll be marinating on your question for some time beyond this blog post, though! Thank you for asking it!

      • Dear Mariana! thank you very much for this detailed answer. I will think about it too. Currently I am writing my PhD dissertation and the choice of words is really important. Thank you for sharing your thoughts!

    • Hi Varvara, interesting point – personally, I tend not to use the notion of ‘wicked problems’ because when taken literally, pretty much ALL real world problems fall within this definition for one reason or another… then the notion makes no sense anymore (and well, if this is reality, not much use complaining that it doesn’t fit our rational problem solving assumptions by calling it ‘wicked’)… Kees

  2. The author does not distinguish between Variable Complex Systems and Complex Systems. This is critical because there are many complex systems such as Engineered Systems which are not variable. That is, they do not involve human decision making.

    • Hi eswonk,
      Thanks for your comment. Indeed, I focus my discussion on human decision-making (ie complex systems / complex problems that involve people and shape people’s lives). I would have thought though that even technical systems service a human end-user? I wonder how often those end-of-the line implications are considered in up-stream engineering. I am a non-scientist/non-engineer myself, so my question is purely speculation.

    • The engineering systems problems you mention probably fall within the definition of ‘complicated’, even if massively so. Complicated problems have solutions that can persist if done well, whereas complex ones do not – in particular, every ‘solution’ changes the problem.

  3. I wouldn’t disagree with Kai’s version of things, and, it appeared to me that the author was focusing on logics AS practical matters more than theoretical ones. As a practical matter, the problem space can turn into a Catch 22 should you not possess the technical familiarity to “solve” the problem. The good part of getting the abduction included, is that everyone – from novice to expert – has some degree of creativity (usually dependent on openness and flexibility) that they can “formalize” into the complexity. After operating programming that focused on complexity, and not just simple to complicated (which is why/where deduction and induction are usually sufficient) for nine years, I’d say that incorporating (formally) more logic-types as a means to learning “co-production” should no longer just be a consideration, but an essential aspect to understanding that solving complexity requires a de-constructive process prior to a reconstruction. Excellent article!!

    • Hello Dean,

      Thank you for your wonderful feedback and comments. I’m thrilled that the intent of my article resonated and came across.
      In my field, working in the national security space we deal with problems that are so diverse: everything from biosecurity, terrorism, drugs, homelessness, environmental crime, cybersecurity…the list goes on. There are many connections across these different fields but what I notice that connects them all is the human-element and secondly, the desire for us (as analysts, policy-makers, decision-makers etc) to feel as though we are in control of these emerging and evolving problems. Being able to embrace the unknown in a way that is not reductive, using our humanness, our lived experiences as well as our expertise, I think opens up a legitimate practice which is often rebuffed (but perhaps shouldn’t be!)

      • From your original post: “Complex problems are open, networked and dynamic.” + “hindsight”.

        To include as an “all of the above” approach, requires a shift in strategy (pragmatic). One from, plan first, to, a plan…that follows prospecting, provisioning, and proposing.

        I’ve found that when dealing with paradox – and in complexity there lurks this other “P” – the logic can be “fuzzy” as in NON-precise, and that can be unsettling for those who want logic to clean up matters. Pierce certainly bumped up against the other pragmatists of his time (William James especially) as there seems to be a bias in human problem-solving for certainty, usually resolved by removing (reductionism) the clutter-in-thinking.

        However, for those on the creative frontier, the boundary is not the container (i.e. start with a case use, start with the plan). It -The frontier/horizon – is the place from which different vectors become tangents, and the one direction of emphasis is NOT chosen…until it is. This is a generative (non-reductive) space and time, not to be confused with the closed and bounded versions of that. Thus, the imperative of reflection, the criticality of hindsight, and the appreciation for the right precursors to planning. Thank you again for bringing this topic to the fore, and no rebuffing here 🙂

        • Thank you, Dean. What you wrote really resonated with me. I’m really interested in seeing how we can bring this type of generative thinking into spaces where analytic thinking and positivist approaches are dominant forms of practice. From what I have seen, experienced and studied, analytic and positivist approaches are really important but their usefulness curbs away when we start looking at these more open problems.

          Another idea that I’ve been playing around with is bias. In the analytic community we work really hard to manage our biases, declare them, wipe them out. However, the very nature of our humanness dictates that there are limits to this. Even AI/machines are influenced by bias based on those that develop them. How might we then rethink this problem as one of how do we step into problem spaces knowing that there will be biases that we cannot eradicate? How do we ‘intervene’ in problem spaces as our authentic, humanly, flawed selves? To be clear, I’m not suggesting we just accept we have biases and move on. Doing so can be quite harmful by perpetuating injustice, power imbalance and vulnerability. I do think that there is this theatre, this performance we do, that makes us feel as though we are in control of all our faculties. We aren’t. What would some alternatives look like?

          Thanks again for a great discussion.

          • One last reply before leaving you to your journey:

            “…analytic and positivist approaches are really important but their usefulness curbs away when we start looking at these more open problems.”

            I think you are pointing to the ‘limits’ problem that arises when things like “combinatorially explosive” and “tractability” issues surface. In other words, when the rate of change goes from incremental to exponential. There is the open-problems type aspect to consider. Otherwise, how what was scaling is now a runaway cascading, wouldn’t be a thing. But it is a thing. A complex thing. So is there a way where the analytics continue, but the data we collect is of a different form (i.e. Bayesian mechanics or quantum mechanics over classical mechanics)? That counterfactual (“what if how we searched for data…changed?”) is something those of us in the Active Inference community, and the Active Inference Institute are asking now.

            “How might we then rethink this problem as one of how do we step into problem spaces knowing that there will be biases that we cannot eradicate?”

            I think it’s a pretty big ask for folks to give up their identity – unless – that shapeshifter comes about because our identity is NOT threatened. To the point you made in your post, abduction begins from a place where the type of analysis needed, assumes that even an expert has to be a novice for a while as they determine the backstory. As a couple of my colleagues point out, Peirce’s schema captures “Reasoning from surprise to inquiry” (Ma & Pietarinen 2016, 2018; LF 3). As such, it portrays retroduction: “given a (surprising) fact C, if A (subjunctively) implies C, then it is to be inquired
            whether A plausibly holds” (Pietarinen & Beni 2021). In a nutshell, this means that what ever “intervention” is to take place, it must take place knowing “we can’t really draw a conclusion until we know more” with a heavy emphasis on the “we” and a nod to “don’t know enough…yet.” When everyone is in that same boat, the differences in the particulars of any one persons identity, gets caught up in the collective identity of common humility (or at least that’s what I’ve found).

            That’s all I have for now. As for alternatives, like the tangents on the margins I mentioned earlier, that’s for folks to self-organize once they realize that they can exercise more optionality when they can incorporate more logics…because they play with those logics (all of them), get familiar with them, and realize that a “leap of faith” embedded in a STEP function, is different than being provided the function requirements then setting about to write the algos that carry that through. Here’s hoping your research (explorations) continue to harvest great insights.

            Ma, Minghui; Pietarinen, Ahti-Veikko (2018). Let Us Investigate! Dynamic Conjecture-Making as the Formal Logic of Abduction. Journal of Philosophical Logic, 47 (6), 913−945. DOI: 10.1007/s10992-017-9454-x.
            Pietarinen, A.-V.; Beni, M. D. (2021). Active Inference and Abduction. Biosemiotics. DOI: 10.1007/s12304-021-09432-0.

    • Hi Jon

      Thank you for sharing this resource. It is great to be part of a collaborative and supportive community. I look forward to digging into it!

  4. The different four qualities of reasoning are extremely familiar to me – but not in that way. I have studied logic decades ago and since then I have continued to elaborate it in my career as a systems thinker. Deductive logic actually can only exist in the realm of mathematics (Gödel, Brink, D. Adams …). Inductive logic is the basis for science: we observe things often enough so we assume it is the general case. If there is no empirical data we have to think of a system that is valid until reality proves otherwise – abductive logic as featured in most models on http://www.know-why.net. A special form of abductive logic is normative logic when we creatively think of how things should be to serve a purpose. And for the latter I have once developed idealized system design (not to confuse with S. Beer’s concept) as an iterative systemic process to find superior solutions. So, quite similar to what you have described and yet different and older (1996).

    • Yes, as a logician I find this article very strange. The author seems to be coining new language for logic, but perhaps the cited authors have already done so. For example deduction is never about causality and induction is only under special circumstances. As for abduction that referred to the important coining of new concepts so as to explain things.

      Moreover, deciding what the outcome should be sounds like action theory, certainly not science.

      • Hi David,

        Thanks for your comments and astute observations. They highlight the differences between disciplines and also how difficult it is to explore logic without risking simplification.

        As you say, I am looking at logic as applied outside the field of science, and a great pick up that I had applied action research to my study (from which my principles for working on complex problems emerged). I used participatory design and co-design approaches to bring disciplines together in my study to explore emerging futures to shape the future we want.

        Do you have alternative definitions of deduction and induction outside the field of science that would be more appropriate, that you might be willing to share? I welcome your expertise in this area! Thank you for sharing your thoughts and engaging in discussion with me.

    • Hi Kai

      Thank you for sharing your rich knowledge and experience in this area. I’m interested in the similarities you mention. It is great to see my thinking endorsed by enduring frames and research. I look forward to following up on the references in your post.

      I also appreciate the differences in disciplinary approaches. Indeed, my framing moves away from the mathematical and science approaches pushing beyond the social sciences into design and creativity. It is useful to think about how these important concepts might present differently when applied to different disciplines (ie origins in maths but evolved for social design or futuring).

      A great discussion.

      • Mariana, thanks for your reply. Indeed an interesting discussion. To support your argument to integrate creativity a short description of idealized system design: Usually we evolve solutions/products/concepts starting from where we are. The alternative is to creatively think of an utopian solution that would probably be unrealistic but super useful. And from that solution we (using a cause and effect model) systematically ask what is needed or what prevents it to find alternative ways. We then end up with a way to realize our original idea or even a completely different idea but anyway it should lead to a better solution than just the continuation of the path we are on. However, many adults find it difficult to think of unrealistic utopian ideas 🙁

        • Dear Kai, we work with similar approaches, however, instead of utopian futures we co-design desired futures with sustainability principles as boundary conditions (to ensure that the futuristic image is sustainable). We often work with the local authorities in Sweden and they say indeed that it is difficult for them to think of such images as they work with short-term thinking on a daily basis: annual budgets, plans associated with 4 years elections. However, business representatives are more keen to such discussions. Additionally, I noticed, that over time such practices get easier for practitioners.

          • Hi Kai and Varvara, it is reassuring to hear that the challenges around encouraging stakeholders to ‘future’ – utopian or not, is a shared experience.

            Varvara, I’d love to learn more about how you’ve used principles as boundary conditions. What does the term ‘boundary conditions’ mean in your context?

            Also, when I read your last line about designing desired futures as a practice that gets easier, it reminded me of the neural pathways we develop when we learn something new. Perhaps, I have to revisit some of my own thinking that some national security practitioners are more operationally oriented whereas others are more strategically oriented. Maybe it does boil down to practice.

            Thank you again for the ongoing discussion, Kai and Varvara.

            • Dear Mariana,

              I can certainly relate to ‘operationally’ vs ‘strategically’ oriented (could you please refer to your work?), in my case, local authorities and planners. Some of them, with a higher position, do have strategic decisions to make, whereas there are planers, who play a smaller role in their everyday work and often do not consider strategic steps to a larger change. Daily work (or broader epistemic standpoint), I believe, doesn’t remain the same during participation in co-production processes (or designing futures as an example), it changes over time due to learnings participants get in the process. Especially, if these processes are iterative. I’ve written a piece with my colleagues called “Lost in translation” (there is a blog post too), where we say that epistemic stand point of participants of co-production processes changes over the course of the process.

              As for boundary conditions, we use sustainability principles (it’s a specific set of principles) to support our co-production processes. It is always a challenge to educate participants of the processes about these principles at the beginning. However, later, when, for example, co-designing desired future(s), we say that this future should not violate the sustainability principles. Furthermore, when creating strategic plans and roadmaps, we assess all the actions against sustainability principles, to make sure that these actions are sustainable.

              You can find all the detailed explanations here:
              Nikulina, Varvara, Johan Larson Lindal, Henrikke Baumann, David Simon, and Henrik Ny. 2019. ‘Lost in Translation: A Framework for Analysing Complexity of Co-Production Settings in Relation to Epistemic Communities, Linguistic Diversities and Culture’. Futures 113 (October): 102442. https://doi.org/10.1016/j.futures.2019.102442.
              Missimer, Merlina, Karl-Henrik Robèrt, and Göran Broman. 2017. ‘A Strategic Approach to Social Sustainability – Part 2: A Principle-Based Definition’. Journal of Cleaner Production, January.


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: