Category Archives: development and complexity

Holism vs. targeting in complex systems

Yesterday, I had a discussion with a friend about the question whether any form of targeting of our interventions towards a specific group of people or topic is already limiting our ability to come up with solutions that are fully adapted to the system. The concrete issues we were talking about were the poverty orientation of the development sector in general and as an example the focus on women economic empowerment as a form of development in particular. The hypothesis we had is basically that if we enter a system with predefined clients (e.g. ‘the poor’ or ‘poor women’) in the first place, our solution will always by biased in order to directly and quickly cater those clients’ needs. This argument goes into the direction of the silo thesis, i.e., that development organizations basically have their topics, such as human health, water and sanitation, markets, etc., and that, regardless of the system or problem they encounter, their solution always has something to do with their topic.

In contrast to that, systems theory tells us that we need to take a holistic view on the system and not limit ourselves to one specific domain if we want to really understand how a system works. So if we only look at the problem of how women get their water and come up with the solution of digging a well in each village, we might miss the whole actual problem women are facing. As a consequence, the entry point for us to support the women might be somewhere completely else, for example local governments, traditional structures, etc.

Then again, what we as development practitioners want to achieve – be it in the short or the long run – is to reduce poverty and also to improve the situation of women, which, in many instances, suffer an even more dire fate as their male counterparts. Not to talk about the incentive structure of the funding of international development which clearly favors quick wins with specific target groups.

I did some more thinking on the topic and I guess the important differentiation we need to make is between what we look at during the system analysis and what we define as target state of the system. In the first, i.e., the system analysis, we have to be open and holistic and take into account all kinds of influences. The only thing we need to do is to set the appropriate system boundaries to frame the system and the level of aggregation that is useful for our work. But only once we have analyzed the system, we should hone in on our target variables, e.g., the poverty or the economic empowerment of women and see how we can influence the system so these variables change in a way that seems favorable to us.

Hence, holism and targeted interventions don’t have to be a trade-off per se. The trade-off, in my view, starts when we are designing the interventions. At that point we have to choose between interventions that have a short-term effect on our target variables or we have to appreciate the dynamics of the system and choose interventions that work with the system. The latter often have the price tag of only showing results after longer periods of time – although that is not necessarily always the case.

I’m curious about your thoughts on this topic and whether it makes sense what I am writing here.

How to plan in uncertainty?

I haven’t been very actively writing here recently. I think I got trapped in the question ‘what meaningful could I write that is not already out there?’. Well, anyway, just a short post today with some thoughts that I have carried around for a while. I just came across a post on the Aid on the Edge blog that was talking about South Africa and the uncertainty of its future:

We human beings do not like uncertainty. We seek to understand what events portend, taking comfort in coming up with an answer. (…) Yet sometimes there is more wisdom, and more comfort to be taken, in acknowledging a more humbling truth – that which of many alternative futures (including ones we cannot imagine) will come to pass is unknowable, is a product of decisions and actions that have not yet been made. This understanding of change as something ‘emergent’, evolving, which can unfold in far-reaching yet ex ante unpredictable directions, is the key insight of ‘complexity theory’ – an insight which can offer a useful dose of humility to governance prognosticators.

The question that comes to my mind when reading this is how to handle the tension of the uncertainty of the future and the deeply institutionalized need for planning in development institutions.

I have worked with a systems dynamics approach combining causal loop diagrams with a method called the sensitivity analysis. It helps us to determine the relative importance of impact factors in a system and characterize them as active, critical, passive, and buffering. Together, these two tools allow to select impact factors that could be targeted by development agencies in future projects.

Now, what is the value of causal loop diagrams? Some people say that they are not more than an improved version of linear causal chains, but still not able to reflect ‘real’ complexity, i.e., the unpredictability of complex systems. Loop diagrams still work with cause and effect relationships, the cause and effect between two factors that can be connected with other factors to eventually build a loop. Yet, complexity sciences say that in complex systems cause and effect are hard to determine, so why bother?

I think that the causal loop analysis and the sensitivity analysis allow us to evaluate the factors that are most relevent and focus on them. They further illustrate some of the more prominent feedback mechanisms of the system that could amplify or hamper our interventions or that we could even use to change some of the dynamics of the system in our favor. They also cater the need of planning or at least of establishing a rational base for planning.

But yeah, we have to avoid falling back into the ‘we can predict the future’ trap, trying to build a prefect model of a system (just remember, models of complex systems have to be as complex as the real thing to accurately simulate it). Complex systems remain inherently unpredictable and our actions need to be tuned to the reactions of the system to any intervention. The above mentioned tools help us to make sense of the dynamics of a system and to select the more promising interventions. They do, however, not release us from the need of an experimental (or may I call it evolutionary) approach to solving real problems in real systems.

I would appreciate any thoughts on that in the comments!