Posted on
April 22, 2024

From an isolated laboratory to a world where "context is everything"

Why is it so difficult to replicate stories of success?

Why is it so difficult to "scale" good solutions when we find them?

Why there are no simple recipes for solving today's biggest challenges?

You might have indulged in questions of this kind before. Why when something goes really well in one situation, it is so hard to apply the learning points from that solution in another context / domain / in another country's offices? And yet we see that a lot of time when we ask this kind of questions, the answers are either a seemingly vague "It depends" which sounds reasonable to some and frustrates others; or a well-meaning "Of course it will work here too" which will fail the moment this assumption exits the realm of textbooks and nice-looking theories and meets the real world.

In this blog post I explore our temptation for “rollout strategies”; then we will take a look at a complexity framework to understand the role of context in different situations; we will see the journey that context has travelled through over the last decades; and how this could have very practical consequences in our work when we deal with complex situations like environmental sustainability, climate change, aid and development.

Let’s start from where this is all too familiar.

The need to “scale up” good solutions is evident (see climate crisis) and the temptation for quick wins and highly replicable solutions is strong in us. Institutions around us are geared to reward this potential to ‘scale things up’, be it an accelerator for start up’s (first question they asked to my consultancy: “how will you scale your product?”) a government grant (“could this be replicated and positively impact more people?”) or aid money (“let’s roll out this new technology in other countries”). Now, the temptation is for a linear pathway made of: testing out something -> looking for what works -> identifying a recipe of sorts -> do a big roll-out implementation. Progress is achieved, the world is changed, we are all happy. This is fine in situations where replication is easy -except that in the most complex scenarios we have to forgo this hope that a copy-paste approach to roll-out solutions be a viable option.

The whole notion of scale implies ‘standardization’ and comes from the idea that you do a replica of your successful project elsewhere. The trick with a standardization is that it minimizes (or better say: bypasses) the importance of context. We get excited about scaling an idea that works: it becomes a trend on twitter and makes the headlines and we all get excited around flashy headlines of the type

“this [new technical solution] could end [systemic problem] in [developing country we are addicted to mention]!”.

Other than some obvious considerations like the lack of a systemic approach to the challenge, at times ignoring the context could be the most troubling of all blunders. This “theory of action” usually works well in a mechanical/technical system, where the level of adaptation to a local context is minimal. The reality though is that when we apply a solution to a rich, complex human system, context becomes everything and the solution we devise needs to be emergent: an answer that is so context-specific that will only work in that context, with at best the possibility to teach us something about a blue-print for actions in other context, but that can never give us a manual with the inspiring headline “here is how you can initiate change in your community!”

To understand more about how context went from “irrelevant” to “everything”, let’s take a walk through a complexity terrain with a brief overview of the Cynefin framework.

Cynefin framework to make sense of systems

Originated during the days of knowledge management and support for decision-making in IBM, the Cynefin framework (a good introduction can be found here) is an approach to understand and act in complexity rooted in (among others) complexity science, anthropology, and cognitive sciences. While you can read a lot about this framework in some good introductory piece like this, for brevity here I will say that you begin with a system / problem / scenario that you need to understand (make-sense-of) and act upon. As a rule of thumb, you want to know “what do I need to know about this system so that I can act effectively in it, to achieve desirable results?” Not all systems are born equal: their nature and their current state can be radically different, and Cynefin suggests that for different current states of the system, different approaches to understand it are required. Let’s take a look. A system can be in a state which is

Ordered: where the links between cause and effect is either known, or in any case know-able with sufficient research and expertise; They are predictable and repeatable. Ordered systems are either Obvious or Complicated, and they require different ways of studying.

Obvious systems are ones in which the link between cause and effect is apparent in plain sight and does not require any additional study or expertise. Here, the decision-making process is to sense the type of situation / problem at hand, categorize it, and respond with the best practice available.

Complicated systems have additional variables and the link between cause (or causes) and effect is not immediately apparent (for instance because it is distant in time and space) but can be known with the right knowledge at hand. Our decision-making here is to sense the problem, analyse it in sufficient depth, and respond with a number of right ways to address it. Expertise here is key, whether coming from an external consultant or by having the right scientific knowledge of the system which will make us predict

Un-ordered: where the links between cause and effect are inherently unpredictable. Either the link is not visible at all, like a chaotic situation, or it is only apparent after the fact (“Oh, that’s what happened, now we get it”). More study of the system won’t bring us to a full picture of it, and even less so to an ability to predict what will happen next.

In Complex systems, the links of causality and the corresponding understanding of what happened can only be seen in retrospect. We can fully make sense only by looking back at what happened; the system does not show itself its initial conditions fully, because information is decentralized, traveling fast, and variables keep evolving, and hence even less so can we predict what will the system do next. Still, the system can show regularity in its patterns, and inclinations to produce certain outcomes (what Dave Snowden calls dispositionalities). When you want to devise a path forward in a complex system there is no way to fully understand its patterns until you start playing around and ‘poking’ the system (because, e.g., only then you will see how rigid its resistance to change is); later you sense the system and respond adequately in a constantly dynamic relationship.

Chaotic systems / states don’t have any apparent link between cause and effect. The system is simply so unstable that requires decisive action to bring it back from the fire and into a safe enough condition where you can properly work with it. These are emergency situations, chasms, moments of panic, etc.

A couple of notes here. 1) This is not a categorization model, but rather a sense-making and support for decision-making framework, hence the four domains are not a way to put systems in boxes but rather a way to help us make sense and ask questions to have a good guess at a) how is the system behaving now? (that now word is crucial) and b) how do we make sense of the system? and c) what do we do now with what we know? 2) the lines in fact are permeable boundaries, and not boxes because 3) generally the systems are highly dynamic and can move from a state to another rather quickly (as we notice when we slip into chaos all of a sudden). Given all this, it is a simplification to say obvious ‘systems’: we could as well say a system ‘currently in an ordered state’, to give an idea of its dynamic nature.

What has all this got to do with our initial musings over development work, scaling up, and context? To understand this, we will turn our attention to an acorn through which a giant of systems thinking gave us a good insight into our recent history of science from linear to systemic.

Russ Ackoff, from the laboratory to the garden

One of the brilliant minds of systems thinking of the last century, who endures respect from a lot of scholars from management to systems science was Russell Ackoff. At once a fine intellectual and a pragmatic management consultant, the Penn State professor in a number of writings and conference speeches elaborated on a little-known but influential paper and explained the relationship between causality and context using an acorn and an oak tree. In the traditional scientific view, Ackoff argues, if A causes B in a linear fashion, then a scientist would say that A is necessary for B to occur, and that A is also sufficient for B to happen. If A happens, B will happen with a probability of 1. This would happen in a world where linear causality is observable, where the connection is ascertained, and where all the other variables have been taken out of the equation. How do you call such a place where you can observe a relationship this straightforward? You call it a laboratory. In the history of science, the laboratory has been designed purposefully to obverse and model the behaviour on natural phenomena in ways that were isolating all the other intervening variables. The lab was a sanctum that got rid of the ‘environment’ where all the noise intervened and muddied the ‘pure’ causal links that happen between A and B. As Ackoff cogently explains though, to call some scientific laws “universal” in the early days of Newtonian physics was a bold statement because they all assumed causal relationships in an empty world with no environment in it.

Take an acorn instead. Ackoff rhetorically asks: what’s the relationship between a seed and a tree? Can an acorn be seen as the cause of an oak tree? Well, it depends. Not in the traditional sense of A will cause B with a probability of 1 (A is not sufficient for B to happen). In an old-fashioned understanding of science, a Newtonian-inspired view of causality posits that you should consider the link between A and B with a probability of 1. But they would say that an acorn had a non-deterministic, probabilistic effect on the oak tree because there were many cases in which you would plant the seed and yet the tree would not accrue. But it is of course necessary to have an acorn in the soil for an oak tree to grow one day -no acorn, no oak tree. What else is needed for a tree to grow? You also need soil, moisture, water, the right micro-climate. In short, you need the right environment for that result to occur. Ackoff refers to an idea developed earlier by philosopher E.A. Singer and calls this relationship between an acorn and a tree the “producer-product” relationship. Ackoff highlights the ramifications of this perspective (source here)

“The view of the universe revealed by viewing it in terms of producer-product is quite different from that yielded by viewing it in terms of cause-effect. Because a producer is only necessary and not sufficient for its product, it cannot provide a complete explanation of it. There are always other necessary conditions, coproducers of its product. For example, moisture is a coproducer of an oak along with an acorn. These other necessary conditions taken collectively constitute the acorn’s environment. Therefore, the use of the producer-product relationship requires the environment to explain everything whereas use of cause-effect requires the environment to explain nothing. Science based on the producer-product relationship is environment-full, not environment-free.”

And environment-full theory of explanation was, in Ackoff’s view, a cornerstone to look at the world through the lens of systems thinking. Its major implications are that when you observe a system as a set of interrelated parts (not a mere collection) that generate a behaviour that none of the parts can generate by itself, you as an observer have to draw its boundaries to conceptually define it. The boundaries are always to an extent arbitrary but you can still opt for heuristic usefulness. Your next step is to look at the system in the nested systems it is embedded in. At a conference speech, Ackoff expressed it thus: (source here)

“Nothing can be understood independently of the environment. What a shocker this was. As a child, I learned there are lots of universal laws, and the first one I learned was that everything that goes up must go down. That’s not true. It’s true within the gravitational pull of the Earth, but go out beyond it and you will go up ad infinitum. Every law is constrained by the environment within which it applies. There is no such thing as the Universal Law. They are all environmentally relative. It was the first consequence of producer-product thinking.”

If you look at the history of science from this perspective, you might have noticed that one of the trends is the increased difficulty in finding out valid applicability of scientific knowledge onto our rich, nuanced, ever changing context. In the best case, this can only happen with a lot of sensitivity to context and adjustments to this landscape’s very specific conditions. In the worst case, the assumption that linear causality must work in this situation as well can actually lead us astray.

The journey that context has traveled

There have been many accounts by now of how our traditional approach to Western science that begun from the Enlightenment -which for simplicity we call Newtonian- is not fully able to take into account the complexity and unpredictability of today’s world. Many have criticized the traditional approach to science of the last two centuries as being ‘reductionist’; usually the word is used as a pejorative term by those who advocate for a big-picture view of reality that takes into account not only the parts (e.g. the leaves) but puts them into larger contexts in which they are nested (leaves belong to a tree, which belongs to a forest, which is created by the micro-climate around them and at times shapes it back, etc).

I choose a specific angle of what has been changing over the last few decades, which is the increasing difficulty to apply linear causality to understand the messy, complex situations the world faces us with.

We can see that everywhere: in our work with development, aid, crime, addiction, and many other arenas. We have solid science that tells us what causal links are at play (for instance: A leads to B) but there is no certainty that a strong correlation found in a study from a social context will extend well into another social context just as well. Say we found that an increased number of firearms in one community is linked to higher rates of violent deaths, and say that such finding has been replicated in many other communities as well. When we take a walk through the states of increasing complexity of a system, from obvious through the complicated to the complex, we will notice how our approach to causality also needs to change in order to help us fully make sense and act in the world. Let’s go back to the lenses we can borrow from Cynefin.

In a system that is in a ‘simple’ state (­a.k.a. obvious), the relationship between cause and effect is immediately apparent, we can zoom in and see how the causal links are at play, and the dependence on the broader context is minimal. We don’t need to know how to fry an egg in South Africa versus in Spain because it’s an obvious task and we can make safe assumptions about gas-powered kitchens the world over and zoom into the stoves, pans, and ingredients, and ignore the rest.

When the system is in a complicated state, we need some more information about the system to analyse it fully and act on it. This will require some adaptation from one context to another: the expert that comes in and researches about your problem in order to “fix it” for you will need to customize her solution, but overall TV cables are TV cables and will fit in with some adjustments with some time and accurate knowledge. Or you will do it yourself, with knowledge that you may not fully have yet but all answers are knowable and a number of good solutions are there to be found. Causal links are evidently at play, but you need to take stock of the environment around you and adjust accordingly to make things work.

When in a complex system, your relationship to context is everything. This is why research on causal links is useful but nowhere near enough to get the system moving towards a different, more desirable direction. This in fact is also a fundamental reason to look at computer-generated models (like the many that agent-based models supply us of late) with a grain of salt when it comes to their explanatory power to simulate and even predict what a complex-adaptive human system is going to do next. (For one thing, because we are nowhere near having an ability to model free will in the agents-which would require a big conversation of its own). The models in fact can never replace nor predict what a complex system will do, nor give us an environment-full explanation. In Cynefin, as we said, true knowledge of a complex system can only arise as we dynamically engage with it and immerse ourselves into its environment by nudging and probing it first.

We know since the time of cybernetics that only variety is able to meet variety, and hence if a problem at hand presents itself with a level of diversity and contextual dependence that is greater than the palette of options included in our latest recipe-book of success stories or our Five-Step-Approach to Successful Change in Organizations, we run the risk of not meeting such variety and applying a sub-optimal solution that backfires (or that shows only initially some minor change but then falters). Why? Because the richness of the context will show us other dimensions, weak signals, an additional Sixth step that was not in the book. And the freer we are from our conceptual baggage of tending to diagnose the system with pre-existing models, the higher chances we have to engage with the complex adaptive system in a productive manner that could accrue long-term results.

So what?

If you have read this far, it’s legitimate to ask, What does this all mean then? These considerations about producer-product, complex systems, and context, all have important implications for how we see the world, and how much we rely on best practice books, expertise, and the “copy-paste temptation” (they have done it there, let’s do it here!)

When you work with a truly intractable problem, the solutions need to be emergent, context dependent, and originated from a dynamic connection to the environment. They will not be known in advance.

One implication is how we relate to our cherished ideas of expertise, and its role in our work. Don’t get me wrong, I am a huge fan of evidence-based, science-informed decision making. Given our current political climate in OECD countries, we need more of that, not less. But such an evidence-informed approach has to be coupled with giving power to the stories of the people and their perceptions of the reality they live in. The more we move from simple towards complex, the more we have to distribute the authority of cognition of the world. Case in point: climate change, fake news, people’s perceptions of something vs the facts that back it up. Education has huge implications and we need to do more of it (especially a serious education of building blocks of epistemology) but in the meantime, we have to work with people’s meaning-making as well else any solution will be perceived as an external body by the cell (and you know what a cell does when an intruder is spotted. It kills it). To truly engage with the context in a social system cannot rely solely on an expert-based analysis of the system (which again is a condition that is in many cases necessary but not sufficient) but has to engage with how the actors in such world make meaning out of what is happening. This is what Kahane calls the “social complexity” layer that is on top of the “dynamic complexity”. This extra layer is made of people’s worldviews, perceptions, and beliefs, and we can’t ignore it.

And now what?

For the decision-makers: you are surrounded by consultants who are eager to tell you about their latest model, approach, solution? Be aware of a) the type of problem you are facing, and the degree of complexity it has and b) how much of their approach is promising a solution that will be applied devoid of context vs how much they listen to the specifics of your own needs and unique circumstances. If the problem is truly obvious or complicated (or presents evident aspects of the ordered domain within it) to rely on best / good practices makes absolute sense; please go ahead and rely on such practice! But for the parts that are truly complex, make sure that an active engagement with the context is indeed happening. Stay open to emergent solutions. Sit with the uncertainty of not-knowing a viable path ahead before you embark if you can. We know, it’s difficult: our minds are quick to identify both causes (even when they are not there) and solutions that may be wrong but sound doable, and close in to convergence before we explore the full range of the issue that we need to tackle. Try and use these simple habits to deal with complexity to help you navigate uncertainty.

For the change-makers and the facilitators: we are torn between the need to “scale up” and roll out solutions (see sustainability, climate mitigation, a more equitable society) and the reality of “how change happens” which, as Duncan Green reminds us, eludes any textbook or best practice. There is no such thing as best practice, alas, even though our public grants and prospect clients keep asking us for certainty about outcomes. I wonder what is our role in this. Shall we be more vocal in saying “to be honest, the world is complex and we don’t actually know how things will turn out”? Blueprints are often fine, heuristics work even better, but there is no recipe in complexity.

Some resources that have informed this piece.