Discover more from Seeds of Science
Prediction and control in natural and social systems
Author: Arturo Macías
Arturo Macías is an Economist at the Spanish Central Bank. He has done some academic contributions on voting mechanisms, natural resource economics, and the mind-body problem, and regularly writes on the Effective Altruism Forum.
Introduction and summary
This essay argues that the exact natural sciences (Physics and Chemistry) are exact because they operate on artificial systems that have been developed in the scientific process itself. Their extraordinary predictive record has created an illusion of general predictability for all natural systems, while in practice, natural science is far less predictive on complex systems. A further distinction is proposed between systems amenable to experimental intervention (modular systems) and systems where observation or natural experiments are the main source of empirical information (non-modular systems).
Social systems are complex and non-modular. Even experimental economics works with subjects from a given culture and social environment, so the ceteris paribus hypothesis (and the external validity of the results) is limited. When social sciences are put in their natural epistemic framework (complex non-modular systems) their degree of success and techniques are in line with those of similar natural sciences.
To some extent system controllability is the main determinant of scientific success. Consequently, the path towards more efficient and controlled social systems depends not only on a more exact understanding of human society but also on social science becoming a constructivist project where society is transformed with governance mechanisms designed to be homeostatic, predictable and welfare optimizing.
The secret of Exact Natural Sciences
The canonical narrative of the birth of modern science describes the development of Celestial Mechanics in the period between the heliocentric model of the Solar System (discovered by Copernicus in 1543) and Newton's discovery of the “System of the World” (1687). The complete description of the Solar System dynamics based on the three laws of motion and Universal Gravitation represents the earliest and most powerful success in the mathematization of natural science. In addition to its own relevance, there is a mythology around this story that highlights the conflict between science and faith (the posthumous publication of Copernicus's work and the trial of Galileo), and the idea of "consilience" (Wilson, 1988), reflected in the history of apple that falls on Newton's head dragged by the same gravitational force that keeps the Moon in orbit.
However, modern Philosophy of Science (both that which descends from Popperian falsifiability or Kuhn’s paradigmatic framework) does not emphasize the most interesting element in the history of Celestial Mechanics: the singularity of its object of study, that is, the Solar System. What makes the Solar System so unique is that it is one of the few, perhaps the only natural system to which the exact natural sciences are applied with almost complete predictive success.
This statement will undoubtedly be shocking to readers with a scientific culture, who have been educated in the idea that there is a relatively small set of mathematically expressible fundamental laws of Physics that explain the entirety of reality. Philosophically, I share that reductionist, emergentist and physicalist worldview (Macías, 2023). However, in practical terms the history of the development of the exact natural sciences (Physics and Chemistry), and the history of technology reveals that the systems for which a high predictive capacity has been achieved in practice are artificial.
The quantitative precision of Newton's laws has been replicated in other physical systems like lenses, batteries and electrical circuits, motors, chemical reactions, or microchips. However, most of these systems (unlike the Solar System) are human creations. Copper wires and batteries where electromagnetism is exactly applied are highly manufactured products. The molecularly homogeneous compounds to which the rules of stoichiometry and chemical thermodynamics can be applied are the result of a long process of discovery and complex manufacturing. The first task for historical chemists was manufacturing the “pure” substances which are the alphabet of their discipline.
Modular and non-modular Complex Natural Systems
Beyond the exact natural sciences, there are other natural sciences: those that study complex natural systems, both animate and inanimate. Within complex systems, a further categorization emerges: one that differentiates between “modular” systems—those amenable to controlled experimental manipulation—and “non-modular” systems, where such direct experimentation is not feasible.
Modern medicine is based on the possibility of carrying out controlled experiments, that is, taking two large and randomly assigned samples of patients, applying to some of them a “treatment” (and not to the other), and describing the differential effect of that treatment. A medication that “works well” does so “on average” in the treatment group, but in general it is very difficult to infer the effect of a treatment on a specific patient. Of course, physicians and biologists develop some causal understanding of “why” some treatments work and others don't, which in turn allows for the creation of more powerful and specific treatments. Even so, the gap in terms of precision and predictive capacity between Medicine on one side and Physics or Chemistry on the other is indisputable.
The last type of natural systems that will be considered is precisely the most difficult to address: the non-modular complex systems. In these systems it is not easy to delimit sufficiently isolated subsystems to be able to carry out controlled experiments, or the size of the systems makes manipulation impossible.
Among non-modular complex systems, we highlight the atmospheric system, complete ecosystems, or populations of a given species (both in the present and throughout Natural History). The natural sciences that study these systems are Climatology, Ecology, Population Genetics and Evolutionary Biology. In these systems, exact mathematical description is impossible, and experiments cannot be “replicated”, either due to the magnitude of the systems or due to the impossibility of approaching the “ceteris paribus” condition. The modularity of a complex system is not a binary characteristic: in most cases there are “natural experiments” or “quasi experiments” (vg. two different islands with a “clear” difference and a different ecology). But is the “clear” difference the only relevant one? Excluding random assignment the equivalence of treatment and control groups is difficult to assure and often controversial, and even more so extrapolating the results beyond the scope of the considered case (the external validity problem).
Lacking the strength of direct experimental manipulation, those sciences address the complexity of their objects of study through a form of mathematical modeling more metaphoric (less descriptive) than that of the exact natural sciences, but no less complex. The Lotka-Volterra equations, Sewall-Wright causal graph models, Maynard Smith and William Hamilton equations for fitness or the general atmospheric circulation models are the products of this modeling efforts, and all of them are more explanatory than predictive.
The previous discussion allows us to classify natural systems and the Sciences that study them into the following groups: i) The simple natural system (the Solar System, studied by Celestial Mechanics), ii) simplified artificial systems (domain of the exact sciences of Nature: Physics and Chemistry), iii) modular complex natural systems (Biology and Medicine) and iv) non-modular complex natural systems (Climatology, Population Genetics, Evolution).
The human society as a complex non-modular system
Social science is extremely wide both in its objects (that rank from the anthropology of hunter gatherers, or the inference of fundamental laws of history, to financial economics) and its methodology (from Hegelian dialectics to Econometrics and Game Theory). Reductionist intent is the most characteristic element of Science (Macías, 2023), and in my view the expression “social science” shall be reserved to the reductionist approaches to the study of human society (reductionism is named “methodological individualism” when applied to human interaction). When reductionism and rationality assumptions are combined, the result is the classical “economic imperialist” approach (Myerson, 1999), where Game Theory is the foundation; by relaxing rationality (but keeping reductionism), the field of “Behavioral Economics” appears.
The epistemic difference between social and natural sciences, which is evident when comparing integrated circuit design with macroeconomics blur almost to the point of disappearing when the discrepancies between Keynesians and neoclassicals are compared with those that divide evolutionary biologists on the issue of “group evolution” (Okasha, 2006). Dynamic Stochastic General Equilibrium Models and Climate Models are reasonably successful in explaining structural relationships, and they run into all sorts of fundamental difficulties when used as a predictive tool (see Muellbauer, 2015 for economics and Fildes and Kourentzes, 2011 in climate modelling).
The analysis of data in complex non-modular systems always lead to the development of a complex statistical theory for causality assessment based on structural equations (Pearl, Glymour and Jewell, 2016), and structural equations are used both by social scientists and biologists (predictably, ecologists are especially intensive users).
Conclusion: the institutional design alternative to High Modernity
Our previous discussion suggests that social systems and their scientific study are not “different” from similarly complex/controllable natural systems. The degree of complexity and controllability of a system is the main determinant of the degree of predictive success of a scientific endeavor (the “hardness” of the Science that studies it). The animistic idea that the human element changes the nature of the social system, and that “social science” is an oxymoron looks quite plausible when the Exact Natural sciences are directly compared to Economics, but when similar natural and social systems are compared, the social/natural boundary becomes less relevant.
Our epistemological classification leads to an immediate practical suggestion: the simplest way to increase the “hardness” of any science is to move from observation to control. Humanity has conquered the natural order mostly by intervention: engineering is based on the development of artificial systems designed to be controlled and stabilized. Artificial homeostasis (Maxwell, 1868) is the core of our technology from the Watt governor to microelectronics.
Recently, there has been heightened interest in governance innovation, driven by factors such as populist politics and advancements in blockchain technology. New schemes for money, voting, public goods funding, sovereignty, and ownership have recently been proposed (Posner and Weyl, 2018; Weyl and Tang, 2024, Srinivasan, 2022; Buterin and Schneider, 2022).
Since Cleisthenes and Solon, even rudimentary forms of institutional design have proven able to transform societies, such that have achieved high levels of individual emancipation and economic and geopolitical success (Ober,2008 and 2015; Macias, 2022) mainly because of institutional superiority.
Unfortunately, after the implementation of welfare capitalism in the postwar period (Judt, 2005), institutional progress appears to have plateaued. The most advanced social systems, both in politics and private sector governance still resemble those of the late 1960s, often those of the XIXth century.
Modern Social Science, especially that based on Mechanism Design, can be the basis of a new wave of institutional progress that can be deployed by firms, governments, and other communities. Unlike the High Modernity project of vertical rule by a technocratic class (Herbert, 2007), institutional design is (and has been since Aristotle and the American founding fathers) an inclusive project where the intellectual elite proposes the rules for the efficient integration of the preferences of the entire demos, avoiding both the autocratic temptation and the alienating paternalism (Krpan and Urbaník, 2021) of High Modernity.
The path towards more efficient and homeostatic social system depends not only on the accumulation of scientific knowledge, but also on social science becoming a constructivist project where society is transformed with distributed governance mechanisms.
References
Buterin, V. and Schneider, N. (2022). Proof of Stake: The Making of Ethereum and the Philosophy of Blockchains. Seven Stories Press
Fildes, R. and Kourentzes, N. (2011).Validation and forecasting accuracy in models of climate change. International Journal of Forecasting
Herbert, U. (2007). Europe in High Modernity: Reflections on a Theory of the 20th Century, Journal of Modern European History
Judt ,T. (2005). Postwar: A History of Europe Since 1945, Penguin Books
Krpan, D., Urbaník, M. (2021). From libertarian paternalism to liberalism: behavioural science and policy in an age of new technology, Behavioural Public Policy
Macias, A. (2023). Freedom under naturalistic dualism, SSRN (see here in Less Wrong)
Macias, A. (2022). El problema de la expansión republicana, Revista de Occidente
Maxwell, J.C. (1868). On governors, Proceedings of the Royal Society
Myerson, R.B. (1999). Nash Equilibrium and the History of Economic Theory,
Journal of Economic Literature
Muellbauer, J. (2016). Macroeconomics and consumption: Why central bank models failed and how to repair them, VoxEU
Ober, J. (2008). Democracy and Knowledge: Innovation and Learning in Classical Athens, Princeton University Press
Ober, J. (2015). The Rise and Fall of Classical Greece, Princeton University Press
Okasha, S. (2019). The Group Selection Controversy in Evolution and the Levels of Selection, Oxford Academic Books
Pearl,J., Glymour, M. and Jewell, N.P. (2016) Causal inference in statistics: A primer, John Wiley & Sons
Posner, E., and Weyl, E.G. (2018). Radical markets: Uprooting capitalism and democracy for a just society. Princeton University Press
Srinivasan, B. (2022). The Network State: How To Start a New Country
Weyl, E.G. and Tang, A. (2024). Plurality: the future of collaborative technology and democracy
Wilson, E. O. (1998). Consilience: the unity of knowledge, New York- Knopf
Technological advances have already made Ecology an experimental science since the early 20th century, including replication, controls, hypothesis testing, etc. It seems like you have confounded the non-realisation of ecological experiments with the quality of the scientific framework, when it is actually due to the absence of means available to do large-scale experiments. Physics has the Large Hadron Collider, etc, because governments made it a political imperative to invest in these types of costly facilities. This has absolutely not happened for ecology. The factor of financial investment is many orders of magnitude different.
Predictability is about having a suitable framework for the object(s) of study and is not a feature of a particular area of investigation. Achieving predictability is, at least in part, technology dependent (e.g. the technology provides an observation in adequation with the hypothesis), as well as being dependent on individual researchers having the means to invest in and use said technology.
The only area of Ecology that has received *any significant* financial investment is in Life Support Systems for space travel, etc - where ecological studies *have shown* that experiments on complex systems are possible, reproducible and predictable and they can lead to identifying conditions of stability. I would just add that the amount of investment remains miniscule relative to any big physical sciences infrastructure. I guess what misses is that Ecology has never really developed a collective focus on a single motivating factor - like understanding the conditions of survival. When Elon Musk realises that he needs a life support system on Earth to ensure his colony on Mars, he will dump money into Ecology. And then I predict Progress Will Be Made.