The curse of evolutionary theories in social science
Theorizing crisis & change in international relations (Part 2 of 2)
I think the way IR theory uses evolutionary arguments and applies them to subjects like institutions or global politics is often misleading. That’s not much of a bombshell but I’m not talking about the usual critiques — that these theories are teleological or that they valorize survival of the fittest. I mean something more fundamental, having to do with the nature of the agents and the way they respond to external pressures.
That doesn’t mean we should immediately discard theories with a whiff of the quasi-evolutionary about them. On the contrary, dissecting the evolutionary metaphor can lead to better-specified mechanisms of social variation and selection. But before I get into the details, let me state my general argument:
1) there is a meaningful distinction to be made between evolution and adaptation when thinking about change in complex systems
2) social processes that we call evolutionary are in fact adaptive
3) this distinction is not semantic or trivial; we should consider it when theorizing change
I’ll begin by outlining some of the differences between evolution and adaptation, then use an example from IR theory to show what happens when those differences are ignored.
When we talk about biological evolution, we are talking about a process of random variation and non-random selection undertaken by actors with no conscious awareness of their task.1 As a result, evolution is fundamentally non-strategic in the Schelling sense. It can neither anticipate nor remember. It can never take a one step backwards, two steps forward approach; it has no memory of the past nor any insight into the future. Because evolution is bound by local maximization, it can be subject to low-level equilibrium traps.
Social processes, on the other hand, are strategic in the sense that actors learn from experience, anticipate moves by others, and plan ahead. “Strategic” does not imply that people are rational or even efficient. We act foolishly, panic in herds, slavishly follow others, and daringly pursue high-payoff/ low-probability outcomes — but we do so with the knowledge of past attempts, and the anticipation of how our actions will be perceived by other actors in the system. (This may be called satisficing rather than maximizing, but even that term assumes consistent behavior parameters that may be absent in irrational or low-information decision-making.)
Unlike their biological counterparts, social processes are rarely evolutionary and often adaptive, in the sense that actors can learn from the past and anticipate the future. It might seems trivially true that conscious agents have a different relationship with environmental pressures than genes do. But the distinction highlights a fundamental difference in the way these pressures manifest themselves in social life.
In evolutionary processes the mechanisms of reproduction and inheritance are discrete — that is, proceeding from generation to generation. In social processes, however, the mechanisms of reproduction and inheritance are continuous, part of the churn of social life. As a result, selection pressures in social systems need not involve the death or reproduction of an agent to manifest themselves. Precisely because actors are aware of systemic pressures, the feedback loop between them remains continuous.
Let me give you a concrete example. Critics of neorealism have argued that state death is too rare to be a plausible outcome of anarchy, and thus no process of “selection” can be observed in the international system. The ‘evolutionary principle’, argues Robert Keohane, ‘can hold only for systems with many actors, experiencing such severe pressures on resources that many will disappear over time’. Bob Jervis, otherwise sympathetic to structural realism, likewise argues that except for a few cases like the Soviet collapse, ‘it is hard to see much natural selection working in this way’.2
Now, we can argue about the merits of neorealism, but this critique doesn’t work on its own terms! In both cases, conflating evolution with adaptation leads Keohane and Jervis to treat state death as the selection mechanism.3 But in adaptive processes, where selection operates through continuous rather than discrete mechanisms, it is not death itself but the ever-present threat of death that acts as the feedback mechanism and spurs officials into action.
In other words, their argument — that state death rates must be high for systemic forces to be powerful — falsely assumes that feedback loops in international politics operate through the mechanisms of evolution rather than adaptation. However, a state in decline feels the pressure of the selection mechanism just as acutely, since the anarchy of politics guarantees, in Stalin’s words, that ‘those who lag behind are beaten’. In effect, the daily execution of government policy provides a continuous feedback mechanism to that state.4 A seemingly trivial distinction becomes important when evolutionary critiques are misapplied to adaptive social processes. Conventional evolutionary approaches to social life mischaracterize the feedback loop through which the environment puts selective pressures on actors.
As Elster puts it, “functionalist social science to a large extent derives from the biological paradigm.”5 He does not mean this as a compliment. There is no reason, however, to treat evolutionary biology as the starting point or the indispensable metaphor for social science. Biology does not have a monopoly on explaining processes of variation and selection. (Malthus’ influence on Darwin serves as the standard reminder here.) Theorizing processes of social change via recurring patterns does not require many of the assumptions we import, implicitly or otherwise, from evolutionary theory. “Cuvier, discussing the observed correlations between ruminants, cloven hooves and frontal horns, has no functional theory to explain this, and is no believer in evolution,” writes Mary Hesse in Models and Analogies in Science, “and yet he takes frequency of co-occurrence to be sufficient indication of the existence of some causal relation stronger than co-occurrence.”6
In sum, I think the difference between evolution and adaptation, in the technical sense outlined above, is worth considering further. My goal is not to offer any definitive statement here but to highlight how misleading it is to think of social adaptation in terms of evolutionary transmission mechanisms. Instead of critiquing the metaphor, we may be better off seeing evolution and adaptation as two distinct domains altogether.
(This is part 2 of 2 on theorizing change in IR theory. It’s based on comments for a roundtable at the 2021 ISA meeting. Part 1, which discusses crisis and agency, is here.)
NEXT TIME: Taking a break from IR to talk about Defining Decadence Down.
The selection process is non-random because the environment “has well-defined criteria for accepting or rejecting any given mutation”. Jon Elster (1983) Explaining Technical Change. Cambridge University Press, p. 50. As he notes, this excludes genetic drift or non-Darwinian evolution.
Keohane, Robert. 1986. ‘‘Theory of World Politics: Structural Realism and Beyond.’’ In Neorealism and its Critics, edited by Robert Keohane, 158–203. New York, NY: Columbia University Press, p. 73. Jervis, Robert. 1997. System Effects: Complexity in Political and Social Life. Princeton, NJ: Princeton University Press, p. 105-7.
Jervis’ argument, as usual, is a bit more sophisticated. In part, he acknowledges that decline can serve as a feedback mechanism but doubts its efficacy some states “gain in many important ways by sacrificing their great power status, and the rise and fall of leading states cannot be understood apart from the internal sources of economic growth.”(1997:107) I see these as compatible with the claims above; adaptive pressures are not constant or consistent but will vary over time, in part due to state actions, and remain something to be explained rather than assumed.
As a speculative aside: in democracies, elections and open deliberation provide another periodic feedback mechanism, which brings us back to theories about the informational advantage of open regimes. Being more sensitive to feedback, they may be more adaptive than their non-democratic counterparts.
Elster 1983:49.
Mary B. Hesse (1966) Models and Analogies in Science. Notre Dame University Press, p.83.