In a recent Scientific American blog post called “Big Data Needs a Big Theory“, Geoffrey West calls for a unified theory of complex systems that will advance our understanding of, and capacity to predict, stasis and change in many domains. Quoting at length:
The digital revolution is driving much of the increasing complexity and pace of life we are now seeing, but this technology also presents an opportunity… With new computational tools and techniques to digest vast, interrelated databases, researchers and practitioners in science, technology, business and government have begun to bring large-scale simulations and models to bear on questions formerly out of reach of quantitative analysis, such as how cooperation emerges in society, what conditions promote innovation, and how conflicts spread and grow.
The trouble is, we don’t have a unified, conceptual framework for addressing questions of complexity. We don’t know what kind of data we need, nor how much, or what critical questions we should be asking. “Big data” without a “big theory” to go with it loses much of its potency and usefulness, potentially generating new unintended consequences.
When the industrial age focused society’s attention on energy in its many manifestations—steam, chemical, mechanical, and so on—the universal laws of thermodynamics came as a response. We now need to ask if our age can produce universal laws of complexity that integrate energy with information. What are the underlying principles that transcend the extraordinary diversity and historical contingency and interconnectivity of financial markets, populations, ecosystems, war and conflict, pandemics and cancer? An overarching predictive, mathematical framework for complex systems would, in principle, incorporate the dynamics and organization of any complex system in a quantitative, computable framework.
We will probably never make detailed predictions of complex systems, but coarse-grained descriptions that lead to quantitative predictions for essential features are within our grasp. We won’t predict when the next financial crash will occur, but we ought to be able to assign a probability of one occurring in the next few years. The field is in the midst of a broad synthesis of scientific disciplines, helping reverse the trend toward fragmentation and specialization, and is groping toward a more unified, holistic framework for tackling society’s big questions.
Not to put too fine a point on it, but I think that agenda is unrealistic.
I agree with West that human social systems are best understood as complex systems in the technical sense of that term (see here). Still, on the possibility of law-like regularities in complex systems that extend to large-scale human social behavior and are usefully predictive, I’m skeptical. It’s hard for me to imagine what those laws would look like, but then I know that my incapacity to understand the universe is not a reliable indicator of the universe’s inherent regularity or intelligibility.
At the same time, I think West’s analogizing to physics and the laws of thermodynamics ignores the single most-important difference between the “natural” sciences and the social sciences, namely, the (in)ability to perform true experiments. (N.B. Humans and their social interactions are, of course, entirely “natural,” too, but these are the terms we conventionally use.) Social scientists can only observe the systems we study; we can’t repeatedly perturb them in specific ways under tightly controlled conditions and see how things play out.
The impossibility of experimentation means we’re never going to be able to see the counterfactuals we’d need to see to make clear and confident inferences about rules or laws. That doesn’t mean we can’t find some robust patterns, but those patterns will never be anywhere near as universal and specific as the laws of thermodynamics.
The fuzziness of our understanding also means that the patterns we do see will have only modest predictive power at best. Those fuzzy patterns will allow us to assess differences in propensities with some success, as they already do now, but they will not lead us to sharply accurate predictions about the timing and details of change.
More important, those patterns themselves will change over time, as the underlying system continues to evolve. As West suggests, the changes that are creating new opportunities for analysis are themselves products of exponential growth in the complexity of human society. It’s an empirical question, I suppose, but I find it hard to believe that the processes which beget conflicts between states in the middle of the twenty-first century—an age of nukes and mega-cities and deep globalization—will resemble the processes that begat World Wars I and II in all but the most banal ways. And, of course, that’s assuming that states in the conventional sense are even still around.