Conflicts & Conversations

There is a particular kind of problem that does not get easier the more you study it. You learn more, certainly. But the problem shifts shape to accommodate what you have learned, so you are never quite ahead of it.

Complex adaptive systems produce emergent behaviour: cascading failures, spontaneous order, tipping points. You see it in weather systems, in markets, in disease spreading across a subcontinent. The components are usually simple. The interactions between them are not, and that is what makes these problems wicked in the sense Rittel and Webber meant.

Wicked problems are problems where every attempt at a solution changes the shape of the problem itself. You cannot solve them the way you solve an equation. You engage with them iteratively, knowing that your intervention is also a perturbation. Epidemiology lives here. We write differential equations as though populations are homogeneous, and then the forecast breaks because people changed their behaviour mid-outbreak.

Erica Thompson frames this well: a model is a metaphor. You say one thing is like another, and reason about the original using the resemblance. The question is not whether the model is correct but how far you can push it before it breaks. A compartmental disease model captures transmission dynamics but has nothing to say about whether the district health officer will act on the output. Every model is a selective representation, and the selection is where the value judgements live.

The part I keep coming back to is what happens when you actually try to deploy these things. Taleb's idea of antifragility is relevant here: some systems need disorder to function well. The best surveillance systems I have worked with are not the cleverest. They are the ones that keep working when data arrives late, in the wrong format, or not at all.

There is also a recursive quality to this work. We build models of systems that include us as actors. The model changes how we behave, which changes the system, which means the model is no longer right. Hofstadter calls this a strange loop. In Post-Normal Science the framing is different but the observation is the same: facts uncertain, values in dispute, stakes high, decisions urgent. That describes most weeks in applied epidemiology.

---

Conflicts & Conversations is what I am calling this space. The conflicts do not get resolved. But there is value in sitting with them rather than pretending they are not there. Model elegance versus field reality. Uncertainty quantification versus the political need for a single number. The desire to predict versus the recognition that prediction in complex systems is mostly structured speculation.

The conversations here are unfinished by design.