A recent article by Duncan Watts in Nature prompted me to structure some thoughts that I had about the challenges in creating and predicting changes in society.
Watts, a physicist by initial training but now a sociologist, asks whether social science should be more solution oriented. As I found recently when attempting to build a simulation model of the diffusion of low energy/carbon innovations, there are many different social science theories that could be applied to how people influence each other and make decisions about whether to adopt a new innovation or not. My intention had been to build on existing approaches but the two primary models used in the simulation literature for representing how influence occurs both failed the face value test of sufficiently describing how influence occurs by missing out key features*. Watts goes further and argues that the two primary conceptual models discussed in the social science literature for understanding innovation diffusion make incompatible assumptions and so cannot both be right. And yet, both continue to be used in research studies.
During the recent Brexit debate in the UK, a leading politician declared that we have had enough of experts telling us what to think because they are usually wrong. He was particularly referring to economists at this point but the bald criticism of all experts was somewhat concerning. We do though know through Tetlock’s work that those ‘experts’ that appear on TV are only marginally better (if at all) than a random prediction. But again, these are mainly economics and politics experts rather than, say, physics or medical experts. The challenge as Watts describes it is that in social science (including economics and psychology), there are many different theories for how any particular aspect of social or economic behaviour occurs and experts usually have their own preference. By contrast, in the physical sciences there is usually broad agreement about the majority of issues particularly those that have been studies for sometime.
Being a mathematician by training I found (and continue to find) it strange that, despite the difficulty, social scientists think that it is OK that they don’t have a clear view on many social issues – and don’t appear to strive to obtain one. There are challenges of course; unlike physical sciences, it is difficult to carry out controlled trials and hard to evaluate the effect of specific social or economic interventions since other factors cannot be held constant and the timescale for social change can be long. Human beings are not as predictable as atoms and the economy and society are complex systems in which effects may be non-linear and emergent but in the real world it is not helpful to have a wide range of competing theories to choose from. Tony Greenham at the RSA argues that economics is not like dentistry (see John Maynard Keynes) and it is impossible to have just a single view of how the economy works rather that it is useful to allow a dialogue between the different approaches to occur. Policy makers may with good reason feel confused about the right approach to take to achieve social change because they have to take action. Ultimately, dialogue has to converge to a decision as to which action to take.
Consequently, Watts argues that some of the social science effort which currently goes into trying to further build theories of how specific aspects of how a social system works would be more usefully focused on solving specific real world issues. I recall sitting in a meeting at a research institute that was trying to decide its research priorities and doodling a 2×2 matrix (I am a management consultant by trade after all!). The two axes were whether the project took the academic field further and whether the work would be immediately applicable. I only discovered sometime later that Donald Stokes got there some 20 years before me in his book Pasteur’s Quadrant. Pasteur’s quadrant is the cell in the 2×2 matrix which both contributes to the academic field AND is also practically useful. This is where Watts’ article suggests more social science effort should be focused though it also discusses the challenges in finding and executing work that fits this brief showing that the shift is not trivial.
In the absence of greater consensus on key issues in the social sciences, I think that Watts’ suggestion of shifting more effort into focusing on specific solutions is a good way forward. And surely better than just refining or adding to the multitude of partial theories which already exist.
Watts makes the arguments much more clearly than I can here and so I encourage you to read his paper.
*Some modellers might argue here that the models have been fitted to real world data and so they have been validated. However, given the number of parameters in each model it is possible to fit both to historic data and so both would appear to be good models. This does not help though when you then want to use the model to assess interventions into the system since you do not really know the underlying mechanism and hence the effect of different interventions.