Book Review: Unsimple Truths: Science, Complexity and Policy by Sandra Mitchell
Unsimple Truths: Science, Complexity and Policy by Sandra Mitchell
Chicago.
University of Chicago Press Paperback $20.00 ISBN: 9780226006628 December 2012
In this book, Sandra Mitchell argues for the importance of understanding complexity when thinking about evidence in policy [1]. Her argument has far reaching consequences for evidence and policy in public health.
We are used to thinking about public health challenges as multi-level problems. Take depression: we know that some people have genes that make them more vulnerable to depression [2]. We also know that childhood experiences, unemployment, poor working conditions, low incomes, domestic violence, substance misuse, recessions and national welfare policies all play a role too [3].
The multi-layered nature of public health problems can lead to confusion if these different influences are seen as somehow competing with one another. Different disciplines have their favoured explanations: economists might see unhealthy lifestyles as evidence of different discounting rates; psychologists might see the same phenomena as evidence of cognitive biases or social norms; sociologists might see them as manifestations of structural influences. These differences include disagreements about what things can count as legitimate ‘causes’ of public health phenomena.
Mitchell argues that, while features at one level of analysis are the product of interacting units at a lower level, these features cannot be predicted from the properties of those individuals. Mitchell uses flocking starlings as an example. The intricate forms taken by a flock of starlings is the product of individual starlings following simple rules about proximity to others in the flock. But the fact that the individual’s behaviour is both influenced by and influences their neighbours means that it is not possible to predict the forms that the flock takes. Results from chaos theory suggest that in complex systems like these, no amount of knowledge about the behaviour and starting points of the individuals would ever be precise enough to predict the overall forms that the flock takes. Mitchell takes the argument further, pointing out that the system-level feature (the pattern of the flock as a whole) could influence the genes of the members of the flock, if it protects the individuals within it from predators.
To take a more public health example, no matter how detailed our knowledge of individual psychology (or genetics, or neuroscience), we cannot expect to accurately predict social patterns of health and illness. Social phenomena cannot be reduced to individual behaviour, and causation flows in both directions. The study of social phenomena complements, rather than competes with, behavioural or genetic explanations of patterns of health and illness.
While Mitchell’s argument may not convince from a pure scientific viewpoint, it has huge practical appeal for making decisions in the real world, with an absence of perfect information or evidence.
Mitchell argues that not all causal factors are equally amenable to experimental analysis. Some genes make a person more likely to get cancer if they are exposed to environmental risks factors but not otherwise [4]. So does the gene (or the environment) only ‘cause’ cancer in some circumstances? Mitchell points out that this idea of causation is quite different from the one we have inherited from physics, where forces like gravity and friction can be separated, and their effects estimated independently.
This has implications for the role of randomised controlled trials, which often aim to isolate the effects of an intervention from its context. Mitchell’s argument suggests that this may not be possible (or desirable) in a complex system like public health. It also suggests that, as Harry Rutter argued in a recent webinar on complex systems and public health, instead of focusing on getting the best possible sandbag, we should focus on building the whole wall.
Mitchell also argues that in systems made of many interacting causes, the effect of changing one factor might not be visible in the overall behaviour of the system. She points out that when genes associated with particular illnesses are inactivated in mice, for around a third of genes the animals appear completely normal. This can happen because the networks of genes involved adjust their activity to preserve the physiological function. This robustness can be seen in health inequalities, which are stubbornly resistant to change. Such stable features are unlikely to be the product of one or two causes: if they were, we might expect to have found and changed them by now.
Finally, Mitchell considers the implications of complexity for policy. She argues that these systems display uncertainties that cannot be quantified. This makes the familiar project of calculating expected outcomes of various policy options impossible. Instead of trying to predict the outcome of each policy option and picking the best, Mitchell argues that we should judge our policies by how well they are likely to achieve a minimum set of outcomes across a range of possible scenarios. And rather than identifying the ‘best’ policy and implementing it wholesale, we should take an incremental and iterative approaches to policy development using feedback to see what is actually happening as a result. This approach has been developed under the name Robust Adaptive Planning [5]. My own view is that this would require much more timely and fine-grained data on public health outcomes than is currently available.
Altogether, Mitchell’s short book is a valuable attempt to produce a philosophy of science that is relevant to current research practice in the biological and social sciences, as well as to the kinds of policy problems that confront decision makers in public health and broader public policy.
Steve Senior is a specialty registrar in public health in Greater Manchester. In previous lives he worked as a policy adviser for the UK Government, and completed a doctorate in neuroscience.
References
1. Mitchell SD. Unsimple Truths: Science, Complexity, and Policy. University of Chicago Press; 2009.
2. Levinson DF. The genetics of depression: a review. Biol Psychiatry. 2006;60: 84–92.
3. Dobson KS, Dozois DJA. Risk Factors in Depression. Elsevier; 2011.