Do these ideas lay a basis for answering the basic question, why does social science matter? They do. It is an unmistakable reality that we are embedded within social and political processes that have enormous consequences for human wellbeing and human suffering. The complex interactions of human behavior contribute both to some of the greatest successes of our civilizations and to the worst failures — persistent poverty, violence, civil wars, climate change, environmental exhaustion, and hunger. We need to solve these “wicked” problems (link), but the interventions we choose need to be as well designed as possible to lead to better outcomes. The tools and methods of the social and behavioral sciences are the best hope we have for guiding our efforts as we strive to solve humanity’s most intractable problems. And it is crucial to keep before us the way that progress in social understanding takes place: through careful and persistent research on a great range of mid-level social processes and patterns of social behavior.
Why does social science matter?
What is the good of social science knowledge? In the natural sciences the answer is often pretty simple and pragmatic: natural science knowledge allows us to predict and control aspects of the natural environment. Physics underlays engineering. That’s not always true, and it’s not the only reason we value research and theory in physics, chemistry, or biology. Some areas of physics and biology offer neither prediction nor control. And surely we also value physics for the abstract understanding it offers for how the world works. But prediction and control are convincing and common reasons that many people would give for valuing natural science.
The situation is different in sociology, political science, and economics. Social science hypotheses and theories rarely give rise to important and usable predictions. The reasons for this lack of predictive power are many. Social ensembles (e.g. cities) are causally heterogeneous and open-ended, so even if we have a strong basis for predicting the behavior of a component system, the aggregate behavior of the ensemble is indeterminate. Social processes are composed of contingent but causally important actions by many agents, leading to path dependency and contingency in the outcome. And even isolated processes (e.g. demographic transition) are often under-determined and path-dependent.
So prediction is very limited in the social realm. What about control? Once we have done some theorizing about a social process or problem, we will often have arrived at some ideas about which factors have a positive or negative influence on an outcome of interest. If we are interested in high school graduation rates or delinquency rates, social science research may advise us that “improving attendance improves graduation” or “neighborhood violence enhances delinquency”. These hypotheses suggest interventions and policy reforms.
That said, the same circumstances that reduce our ability to predict social outcomes also limit our confidence in the efficacy of given interventions. The impossibility of designing reliable strategies for regional economic development or job creation are cases in point. If graduation rates are influenced by a handful of inter-related causal circumstances, it is difficult to have a lot of confidence in a single-factor policy reform.
Where the social and behavioral sciences can offer the greatest confidence is in several important areas. Demography is one; the ability of demographers to forecast population size and composition is extensive. Discovery of behavioral patterns in a variety of circumstances is another. Education scholars can learn how children commonly respond to this or that feature of the classroom environment; sociologists can learn how drug offenders usually respond to various treatment programs; rural sociologists can discover how small farmers respond to the availability of new seed varieties. And political scientists and social network specialists can work out the system tendencies of various voting systems or communications network architectures. So there is no shortage of reliable, useful results in the social sciences. What is more problematic is the idea that these sorts of findings might add up to a body of knowledge that permits predictions at the aggregate level of complex social ensembles.
We might say that sociology or political science offer some guidance about population and behavioral tendencies, system characteristics, and rough estimates of the causal properties of various social structures. And we can use these hypotheses and theories to make some informed guesses about the likely direction of change that will result from a given intervention or environmental change. But we are forced always to concede that our expectations take the form of bounded ceteris paribus predictions rather than confident and reliable point estimates of outcomes. So policy formation and social science research have a looser fit than engineering and physics.
I think this assessment fits pretty well with the idea that the social sciences are at their best when they focus on identifying and documenting a range of social mechanisms and processes, or when they pursue Merton’s ideal of “theories of the middle range.” And this leads us to recognize the limits that surround the possibility of aggregating these kinds of mechanisms into confident assertions about the behavior of the large social ensemble. Social policy is not an exact science.