Brit. J. Phil. Sci. 63 (2012), 1–26
Why do Spatiotemporally
Restricted Regularities Explain
in the Social Sciences?
Alex Rosenberg
ABSTRACT
Employing a well-known local regularity from macroeconomics, the Phillips curve,
I examine Woodward’s ([2000], [2003]) account of the explanatory power of such historically restricted generalizations and the mathematical models with which they are sometimes associated. The article seeks to show that, pace Woodward, to be explanatory such
generalizations need to be underwritten by more fundamental ones, and that rational
choice theory would not avail in this case to provide the required underwriting.
Examining how such explanatory restricted regularities are underwritten in biology—
by unrestricted Darwinian regularities—provides the basis for an argument that
Darwinian regularities serve the same function in human affairs. The general argument
for this claim requires, inter alia, that we accept some version or other of a theory of
memes. The article concludes by clearing the field of some prominent objections to the
existence of memes, and extracting some policy implications from the persistence and
acceleration of arms races in human affairs.
1
2
3
4
5
6
7
Introduction
Invariance as Evidence or Diagnosis: An Example from Woodward
Why are Restricted Regularities Explanatory?
Invariance and Arms Races in Biology
Restricted Regularities, and Arms Races in Human Affairs
Dealing with the No-memes Objection to Darwinism about Human Affairs
Conclusion: A Moral for Institution Design
1 Introduction
All regularities in social science are local—that is, spatiotemporally restricted.1 Moreover, mathematical models in the social sciences that explain
1
A regularity is restricted if it mentions specific times and/or places, or contains ceteris paribus
clauses. The latter day locus classicus of the view that all the laws of the ‘special sciences’ are thus
ß The Author 2011. Published by Oxford University Press on behalf of British Society for the Philosophy of Science. All rights reserved.
doi:10.1093/bjps/axr014
For Permissions, please email: [email protected]
Advance Access published on June 10, 2011
2
Alex Rosenberg
do so usually because the domains to which they are applied realize spatiotemporally restricted regularities that approximate these models. It is not
generally held that these spatiotemporally restricted regularities can be expected to give way to less restricted or ultimately unrestricted regularities.
Nevertheless, the spatiotemporally restricted regularities are treated as explanatory ones. This raises the problem of why such regularities are explanatory. What is the difference between them and spatiotemporally restricted
regularities that are accidental and have no such explanatory power.
(Hereafter, I drop the ‘spatiotemporally’ qualifier to ‘restricted’.)
In this article, I offer a surprising answer to this question, one which underwrites a Darwinian approach to understanding human affairs across the social
sciences. My argument will employ a well-known example of such a restricted
regularity from economics, and a proposed answer to the question of
why local regularities are explanatory that is due to James Woodward
([2000], [2003]). Reasons for focusing on his answer to the question include
its contemporary currency, widespread acceptance, and most of all, the deep
study of regularities from which it emerges. Woodward’s approach either
takes account of or subsumes most of the other approaches that have eventuated in affirmative answers to the question of why restricted regularities
explain.
If my example is typical, and if the problem identified for Woodward’s
approach is irremediable, then both the example and the difficulty provide a
powerful argument for treating restricted regularities as the outcomes of
Darwinian processes operating in human affairs. This conclusion in turn
requires me to face obvious objections to a Darwinian approach, objections
often framed in terms of skepticism about whether there are cultural replicators—‘memes’ of the sort required for Darwinian processes in human affairs.
However, I argue, the explanation of why restricted regularities in human
affairs are explanatory also enables a Darwinian approach to deal with
doubts about whether there are ‘memes’ or other replicators of the sort a
Darwinian process requires.
2 Invariance as Evidence or Diagnosis:
An Example from Woodward
The ‘Phillips curve’, is a regularity about the relationship between inflation
and unemployment: there is an inverse relationship between the rate of change
restricted, but nevertheless explanatory is (Fodor [1974]). I assume without argument that the
regularities of physical science are not similarly restricted. If they are held to operate only in this,
among many universes in the multiverse, or to be evolving in this one, suitable qualifications to
my claim can be added.
Restricted Regularities in Social Sciences
3
in money-wages and the level of unemployment. This claim, famously made
by A.W. Phillips ([1958]) in the late 1950s provides a basis for Keynesian fiscal
and/or monetary policy, in particular that unemployment can be lowered by
inflationary government action, for example, interventions such as increasing
the money supply or incurring a budget deficit. Woodward’s approach to
why the Phillips curve was explanatory turns on the nature and role of such
interventions.
According to Woodward, for the Phillips curve relationship between unemployment and the rate of increase in money wages to explain the original
data which supported it and subsequent phenomena to which it was applied,
the regularity about their relationship must satisfy the following requirements
([2000], p. 201):
M1. The intervention (in this case, for example, increasing the money
supply) changes the rate of inflation from what it would otherwise be.
M2. The Phillips curve generalization asserts that increasing the rate of
inflation really does reduce the unemployment rate from what it would
otherwise have been.
M3. The increased rate of money supply growth lowers the unemployment rate only through its impact on the inflation rate, and not by any
other means—for example, by (improbably) making workers reduce their
labor supply in protest at the rate of money supply growth, and thus
reduce the rate of unemployment.
M4. The increase in money supply is not correlated with other causes of
unemployment increases besides inflation, so that we can exclude the
possibility that money supply increases and inflation are both the joint
products of some other independent change. This condition rules out the
possibility that some other variable may work independently of either
one to change the rate of unemployment.
By the 1960s monetarist economists, following Milton Friedman ([1968]),
were challenging the Phillips-curve regularity as an explanatory one. They did
so by denying M2 on the basis of another regularity deemed by them to
be invariant, that agents choose rationally. (Friedman did not deny M1.
Presumably he held it to be an accidental regularity, or perhaps the result of
an erroneous inference from inadequate data.)
Friedman held that in the long run, the rate of unemployment was constant
for any rate of inflation, that insofar as the Phillips curve regularity obtained,
this was a short run phenomenon, and that after a temporary increase, the rate
of employment returns to its ‘natural rate’ as economic agents realize that the
price changes are not going to continue at the temporarily accelerated rate.
Friedman thus sought to undercut the Phillips curve’s explanatory power by
arguing that it is not invariant under changes in the beliefs of economic agents,
employers, and workers. And this failure is due to another invariant relationship, one that figures in rational choice theory. Since rational employers and
4
Alex Rosenberg
workers recognize that changes in money prices may not result in permanent
changes in relative prices, such changes by themselves can have no lasting
impact on their choices.
But the invariance Friedman invoked to undercut the Phillips curves’ explanatory power is itself subject to an objection of the same kind. Experiments
in cognitive science, for example, provide evidence that the choices of actual
economic agents persistently respond to changes in money prices, even when
they are acquainted with the relevant economic theory that vindicates the
rationality of indifference to changes in money prices which leave relative
prices unchanged (Kahneman et al. [1982]). There is both anecdotal and experimental evidence that suggests such a conclusion. Of course such experimental data, if replicated and otherwise reliable, would lead one to conclude
that the regularities of rational choice theory were not invariant in the way
that Woodward requires explanatory generalizations to be. That is, the claims
of rational choice theory must be denied (at least some) explanatory power,
and/or they may not provide a basis to deny the invariance, and thus the
explanatory power, of the original Phillips’ curve relationship.
At this point in the dialectic among economists, the late 1960s, defenders of
the Phillips curve generally insisted that the statistical and anecdotal historical
evidence for the Phillips’ curve outweighs an argument like Friedman’s appeal
to rational choice theory. They had grounds for doing so: first, the continuing
evidence that the curve held; second, they could reject rational choice theory’s
alleged invariants, invoking the operation of ‘animal spirits’ to which Keynes
himself adverted in one of his digressions into informal micro-theory in
The General Theory.2 Both of these considerations gave continued life to the
Phillips curve and its explanatory force through the 1960s and early 1970s.
For, by this time the Phillips curve had been taken up by Keynesians as empirical evidence for the applicability of the Keynesian models of the economy.
Because Keynesians did not give much credence to rational choice theory nor
seek microfoundations for their macroeconomic generalizations, they paid
little regard to Friedman’s argument, and continued to treat the Phillips
curve as an explanatory invariance.
However, by the 1970s increases in inflation were no longer in fact decreasing the unemployment rate. This period was characterized as one of ‘stagflation’ since employment was stagnant while the rate of inflation was increasing.
2
In the General Theory, Keynes wrote only a little about this notion: ‘a large proportion of our
positive activities depend on spontaneous optimism rather than mathematical expectations,
whether moral or hedonistic or economic. Most, probably, of our decisions to do something
positive, the full consequences of which will be drawn out over many days to come, can only be
taken as the result of animal spirits—a spontaneous urge to action rather than inaction, and not
as the outcome of a weighted average of quantitative benefits multiplied by quantitative probabilities.’ (Keynes [1936], pp. 161–2).
Restricted Regularities in Social Sciences
5
At this point, owing to the failure of accelerating inflation to be correlated
with reduction in unemployment, economists began to seek an explanation for
its failure to be invariant. In particular, they sought it in ‘microfoundations’
for macroeconomic models. Robert Lucas ([1972]) and other rational expectations theorists argued that economic models intended to be used by government in policy guidance had to incorporate, or even be grounded on, the way
in which policies affected individual economic agents, and how they would
respond to them. Keynesian models self-evidently lacked microfoundations.
The ‘Lucas critique’ and the rational expectations theories based on it were
consistent with the demise of the Phillips curve and explained why it should
have ceased to be invariant in the period from the 1970s onward.
It is important to note that the Lucas critique gained currency only after the
empirical disconfirmation of the Phillips curve by 1970s stagflation. It was
then that economists began to seek explanations for a failure of invariance
already recognized. Woodward says, ‘The interesting question for economists
is not whether the Phillips curve is a law of nature, but rather whether it is
invariant under certain specific kinds of changes and interventions’ ([2003],
p. 221). But the order of events in the process of abandoning the Phillips curve
did not begin with an initial discussion of the conditions under which its
invariance might fail. Rather the Phillips curve was treated as disconfirmed
by data, and only then did economic theorists set about seeking an adjustment
in economic theory that would explain the disconfirmation. Lucas found one
by re-asserting the importance of microfoundations and re-asserting the rationality of economic agents: in general they take into account policy changes
in rationally determining future actions.
Woodward writes, ‘judgments about the significance or importance of the
intervention over which a generalization is invariant [. . .] play an important
role in the construction and assessment of explanatory generalizations’ ([2003],
p. 221, emphasis added).
This does not appear to have been the case in the construction or assessment
of the Phillips curve. Consider its construction first. Phillips offered his
curve after undertaking the study of a correlation, one which Keynesian
theory might lead us to expect. But in many of the cases included in
Phillips’ data, governments did not intervene to increase money prices, and
the price rises were due to a variety of unintended economic causes. Phillips
was not identifying interventions when he constructed the curve.
As for assessment, the stagflation of the 1970s provided empirical disconfirmation of the Phillips’ curve by data, not interventions. In retrospect it was
clear that the failure of policy-interventions to increase employment which
assumed the Phillips curve correlations did lead to its reassessment. But governmental intervention played little role in the reassessment of the Phillips
curve as an explanatory generalization. The Lucas critique identified quite
6
Alex Rosenberg
different non-governmental interventions as significant in its assessment of the
failure of the Phillips curve as an explanatory generalization. The explanatory
failure of the Phillips curve was traced to its failure to be invariant over interventions in rational agents’ information. It was the failure of the Phillips curve
to hold over these interventions that rational-expectations economists turned
to, after the Phillips curve’s explanatory power was called into question by
good old statistical evidence of correlations, independent of whether these
changes in money prices were ‘interventions’.3
However, there was a serious problem with the diagnosis of the Phillips
curve’s failure under interventions in information of economic agents, offered
by the rational expectations theorists. It is now well-known from the work of
(Kahneman et al. [1982]) and many others that economic agents are far from
rational in many respects. These respects are so many that we can hardly have
much confidence that the invariance of the components of rational choice
theory correctly diagnoses the failures of the Phillips curve to be invariant.
If the regularities of rational choice theory themselves lack invariance under a
range of interventions to which they are applied, then it becomes problematical to employ them in an assessment of the failure of the Phillips curve to be
invariant. In ([2000], p. 232–3), Woodward himself sketches considerations
that have led economists and political scientists to doubt the invariance of
rational choice theory under relevant changes to its variables. Under these
circumstances one must consider how it can reliably explain why the Phillips
curve fails to be invariant.4
The problem this history of the Phillips curve raises for understanding how
explanations work in economics (and all the special sciences, I hold), is this:
The Phillips curve actually explained the relationship between the rate of
change in money prices and employment over some period up to the 1970s.
This period may even have been a long one, going back to the beginning of fiat
currency or even earlier if currency debasement and vast new discoveries of
gold and silver result in inflation. During the whole time it obtained and really
did explain how inflation enhanced economic activity, including employment,
the Phillips curve could have broken down had economic agents’ expectations
about changes in the rate of inflation been different. Had workers not suffered
from a money illusion, had businesses not realized that the resulting higher
3
4
If intervention is synonymous with change, then of course the disconfirmation of the Phillips
curve was the result of noticing that interventions on (changes in) money price levels ceased to
co-occur with changes in levels of employment. Here, I assume that in macroeconomics, interventions are policy-directed changes.
Experiments of the sort (Kahneman et al. [1982]) and others have run do not directly address
‘money illusion’ phenomena, but they do show that there are asymmetries in the dollar valuation
of goods depending only on whether the subject has ownership or not. So, there is good reason
to think that economic agents persistently violate the strictures of rational choice theory, contrary to the assumptions of the Lucas critique.
Restricted Regularities in Social Sciences
7
interest rates would reduce the attractiveness of increased production and
increased employment, had savers not realized that their real savings rate
was too low as inflation increased, in general, had economic agents had different expectation (or as micro-founded macroeconomic theory requires,
acted as if they did), the Phillips curve would have ceased to hold earlier.5
There was no causal barrier to such realization throughout the whole time
the Phillips curve held good. But if the Phillips curve regularity could have
broken down at any time, was it not merely an accidental regularity lacking
explanatory force? ‘Surely not!’ we might want to say.
But consider, the Phillips curve does not support the counterfactual that
even had people realized how inflation impacts their rational choices, changes
in the inflation rate would still be followed by changes in employment levels.
Once people actually realized how inflation impacts on rational choices, the
Phillips curve ceased to have explanatory power. But apparently they could
have come to this realization at any time during the period the Phillips curve
held. So here is the question: why should the Phillips curve ever have had
explanatory power, given that it could have been made false at any time prior
to the stagflation of the 1970s? Why is it explanatory even though it lacks
counterfactual support? This is just a version of the question of why restricted
regularities are explanatory.
3 Why are Restricted Regularities Explanatory?
Woodward writes,
It is perfectly possible for a generalization to be invariant only for
changes and interventions that occur within a limited spatial or temporal
interval and to break down outside that interval. Suppose that, contrary
to actual fact, the Phillips curve turned out to be invariant under
government interventions that changed the inflation rate between say,
1870 and 1970 in the United States, although not invariant outside this
interval. If this had been the case, then (I would claim) despite the limited
spatio-temporal scope of this relationship, one could appeal to it and to
the fact that the US government intervened to raise the inflation rate in
1915 to explain why unemployment fell after the intervention. More
generally, in contrast to the traditional law based account of explanation,
the notion of invariance allows us to talk about explanatory relations
that hold only over limited spatio-temporal intervals or which make
reference to particular objects, events or processes [. . . ] Many explanatory generalizations seem to have exactly these features and this is one
5
It was not merely a short-run money illusion of the sort Friedman granted, nor yet simply a
long-term one, as the Keynesians seem to have held, but a variety of regularities about expectations and behavior that drove the Phillips curve and whose break down resulted in the Phillips
curve ceasing to hold. I owe this insight to Don Ross.
8
Alex Rosenberg
reason why the notion of invariance is particularly well suited to
understanding their character. ([2000], pp. 224–5)
As Woodward rightly says, we treat restricted regularities as invariant and
therefore explanatory, even when they could have been made false at any time
during which they held. For example, in the case of the Phillips curve, ‘interventions’ drawing economic agents’ attention to the differences between
money prices and relative ones could have been undertaken at any time
before the 1970s. It should then have had the same effect that the Lucas
critique alleged it had after 1970. My question is, given this possibility, why
do we treat restricted regularities of economics or the other social sciences as
explanatory?
Compare the reasons we treat the regularities of common sense- or every
day- or folk-physics as explanatory, in spite of their restricted invariance. In
the case of a variety of folk-physics regularities about specific objects or
processes, such as that iron sinks, wood floats, free falling bodies move
toward the center of the earth, it is fairly easy to check for invariance by
experiment, either casual or systematic. Regularities that pass such tests are
accorded explanatory power in ordinary life and they are the starting point for
refinements that systematic science seeks. This is true even when as in many
cases the folk-physical explanations of these regularities are wildly wrong
(e.g. non-zero velocity requires the action of a continuously applied force,
all velocities are additive, space is anisotropic).
Suppose as seems reasonable that the regularities of folk psychology are the
starting points for explanatory regularities in social science (think of
Thucydides). Our confidence in them is not similarly the result of informal
experiments that identify the range of variations in causal variables over which
they are invariant. Rather, the explanatory folk-psychological regularities we
begin with in social science (e.g. if you desire d and believe that action a is the
best way to secure d, you do action a) are supported by some combination of
introspection, the meanings of words, and a bit of informal conceptual
analysis. It is hard even to know how to set about testing folk psychological
regularities for the range of values of the causal variables over which they are
supposed to be invariant. Almost all tests of invariance advanced in these
cases are either indecisive because the experimental set ups don’t replicate
the regularity in question, or else their outcomes are subject to alternative
interpretations, or else replication is impossible or otherwise deemed to
be irrelevant to the explanatory force of folk psychological regularity.
(See Rosenberg [2007] for an introduction to problems facing the treatment
of folk psychological regularities as explanatory.)
Empirically driven social science adds some explanatory regularities to
those of folk psychology, and also provides some formalization of the
Restricted Regularities in Social Sciences
9
folk psychological regularities it adopts. The evidence for these regularities is
often statistical, econometric, or (increasingly) based on data mining. It is
sometimes anecdotal, and rarely based on laboratory findings, controlled experiments, quasi-experiments, or natural experiments. But none of these regularities comes up even to the level of predictive reliability or explanatory
power of many of the restricted regularities of folk physics and folk biology.
In the absence of unambiguous experimental or observational results,
regularities in social science cannot help themselves to the grounds that
make explanatory the temporally restricted regularities of folk physics or
folk biology. So how can a regularity of economics that is true only within
a certain period really be judged invariant in that period and so explanatory
during that period?
Here is one answer that vindicates Woodward’s approach. It actually requires some claims that the passage seems to deny. First, the Phillips curve
probably was invariant for government interventions that increased the rate of
inflation in the period 1870–1970.6 (Woodward’s use of the counterfactual,
‘If this had been the case’ suggests he disbelieves it.) On several occasions
during this period government action increased the rate of inflation, for
example, during the panic of 1893, during the First World War, after the
mini-depression of 1937, and in each case unemployment decreased.
Of course the government intended neither action—increasing the rate of inflation or decreasing the unemployment rate (except in the 1937 case)—but its
interventions had the effect the Phillips curve describes.
The Phillips curve was invariant over this time, because governments were
not persistently intervening to increase the inflation rate, and when they did
so, e.g. by issuing greenbacks during the civil war, the intervention was comparatively infrequent. Thus, economic agents did not have occasion regularly
to incorporate in their economic calculations the fact that money price levels
were changing in ways that did not reflect relative prices. Accordingly, the
increases in the inflation rate were followed by increases in employment as a
Keynesian theory would suggest and as Phillips discovered. After the publication and dissemination of Keynes’ General Theory of Employment, Interest
and Money ([1936]) government policies began to exploit the Phillips curve.
But after 1970 economic agents began to respond differently to inflation’s
persistence, which resulted in part from governments’ willingness to exploit
the Phillips curve relationship. At this point, owing to expectations that inflation will persist and accelerate, the Phillips curve disappears—the life-times of
the short run curves get shorter and shorter, and adjustment to the long run
vertical curve comes faster and faster until we have stagflation, and the impotence of deficit spending as well as expansionist monetary policy. In short,
6
See (James [1989]).
10
Alex Rosenberg
the Phillips curve explains changes in employment up to 1970 because before
then it really was invariant, given economic choices, public policy-choices in
the case of the government, and private policy choices in the case of individuals. Changes in private policy choices responding to public policy deprived
the Phillips curve of its invariance.
If this account of the matter is right,7 the change in the Phillips curve from
being invariant, and therefore explanatory up until 1970, to no longer obtaining after 1970, is the consequence of what evolutionary-game theorists will
recognize as ‘arms races’: strategic interactions between two or more players,
in which each searches for strategies that will attain aims which are incompatible, and in which temporary equilibria between strategies in play are
broken by the implementation of a new strategy by one side or the other, in
a persistent cycle.8 In an arms race an equilibrium among strategies will obtain
for a period of time, sometimes a very long time, and then will be broken, by
one party finding a better response to the strategy of the other. In the USA, the
Phillips curve is the product of strategic interactions over a long period, one
that began when, for example, fiat money was introduced in the USA with the
introduction of ‘green-backs’ during the Civil War, in 1862. Apparently it
required 70 years or so before governments began to exploit this regularity
in its own full employment-stimulation strategies, and another 30 years before
economic agents recognized the impact of government monetary strategies on
the outcomes of their economic choices, and began to take action based on this
recognition which frustrated the government’s strategy.
So, now we have an explanation for the explanatory power of a spatiotemporally restricted invariance: it is a temporary local equilibrium among two or
more players. The equilibrium is temporary because one or more of these
players will sooner or later hit upon a new strategy that enables it to exploit
the equilibrium strategy of other players, thus destroying the equilibrium,
breaking the temporarily invariant regularity it sustains. If we can treat explanatory regularities in social science in this way, we will have an attractive
argument for Woodward’s conclusion that even when spatiotemporally
7
8
It is fair to say that this is the received view about macroeconomic history among many economists and Woodward accepts it (personal communication). Some economists add complications
to the story, ones explicitly involving not just consumers and workers, but also suppliers and
investors making decisions involving opportunity costs of production versus lending, along with
central bankers making decisions about interest rates and the money supply independent of
government fiscal policy, all seeking to exploit several distinct related local equilibria reflected in
the Phillips curve.
I employ the terms ‘strategy’, ‘arms race’, ‘player’ here with the meanings common to evolutionary game theory, which does not require that the players be intentional agents knowingly
employing strategies they can express. Hawks and doves can play hawk/dove without knowing
it, so can governments and citizens. ‘Expectations’ should be understood similarly. For an
accessible introduction, see (Dixit and Skeath [2004]).
Restricted Regularities in Social Sciences
11
restricted, some regularities have explanatory power, owing to their temporary invariance.
If this argument is right, it turns out that restricted regularities are invariant
and therefore explanatory because they are underwritten by less-restricted
invariances, and these in turn by others, until we reach bedrock with strict
laws of the sort which Woodward certainly does not think we need to back up
explanations in social science.9 To see why this is so, we need to turn to the
home base of arms races, biology, and consider how restricted regularities can
be explanatory in that discipline.
4 Invariance and Arms Races in Biology
It is widely held among philosophers of biology that there are no completely
invariant regularities in biology except for those reported in the Darwinian
theory of natural selection. A variety of arguments have been offered for this
consensus view, of which perhaps the most influential is John Beatty’s ([1995])
‘evolutionary contingency thesis’: all other regularities in biology—from the
most invariant to the least—obtain only as a result of the operation of natural
selection on initial conditions that have obtained in the history of the Earth,
and are subject to abrogation by the operation of natural selection on later
conditions.10
The general argument is obvious, and it has immediate implications for
biological arms races. Since nature builds adaptations by a process of environmental filtration of random variations, when environments change adaptations can become maladaptations and vice versa; variations neutral in fitness
in one environment can become adaptive or maladaptive in another one. But
nothing is forever: even the most stable environmental conditions will sometimes change, and even quickly. Consider how the asteroid impact at the
Cretaceous-Tertiarty boundary 65 million years ago changed the environment
and killed off all dinosaurs within a few years. Thus no regularity thrown up
by the process of natural selection is immune to breakdown as a result of
environmental change.
Once the evolutionary environment comes to include creatures and their
effects on one another, the life-times of regularities about creatures’ adapted
9
10
That there will be strict laws at some deep level or other underwriting and explaining restricted
regularities is not disputed by Woodward (personal communication). But strict laws of physics,
say, on which all other processes, including those treated in social science, supervene, will not do
the work which I will argue Woodward’s theory requires. No one supposes that physical laws
have any significant role in social science’s explanations! Woodward’s claim is that spatiotemporally restricted regularities in social science are explanatory even in the absence of any less
restricted regularities, still less strict social scientific laws to underwrite them.
Some philosophers of biology even deny that any invariant regularities operate in evolution (e.g.
Kitcher [1993]). Beatty’s argument and conclusion can be rephrased to accommodate this
denial.
12
Alex Rosenberg
traits fall from the scale of billions of years (Archaea—whose environment
has not changed for 3 billion years) to multiple geological epochs
(oxygen-respirators) to hundreds of millions of years (vertebrates) to weeks
and months in the case of others (the AIDS virus). These evolutionary interactions between biological lineages are the original ‘arms races’. Owing to the
role of environmental change, even the most established and long-lasting
regularities in biology are not as invariant as any well-established regularity
of physical science.11
Given the slowness of most environmental changes, regularities about
individual species can remain invariant over geologically long periods.
Changes such as the shift to an oxygen-rich atmosphere, or continental drift
or the onset of ice ages will break up some invariances and create new ones.
Other, more rapidly occurring species-making or changing processes such as
earthquakes or major droughts will have similar results. But the invariances
produced will be hard to break down for the same reason that massive and
long-lasting environmental change was required to put them in place.
However, matters begin to change quite radically once any species becomes
part of the selective environment of another species. In these circumstances,
regularities about the latter species become comparatively shorter lived, more
temporary and more spatially restricted as well. The reason of course is that
species become part of one another’s selective environments when they compete, or one predates upon or parasitizes the other, or both are predated by a
third species, or both predate a third species, or both cooperate for that
matter. Under all these circumstances, nature is persistently searching through
design space, seeking variations in both species that will provide them with a
selective advantage over the other. Until it finds one, the two species are
locked into a local equilibrium, one which is reported in a temporarily invariant regularity. The theory of natural selection however assures us that, if it
11
Consider what was until recently thought to be the most invariant of biological regularities: all
genes are composed of DNA. For a long time this regularity was subject to no exception. But
because it remained invariant over a very long period, its operation provided an environment
that would allow for the selection for any new biological system that could take advantage of the
fact that all genes are composed of DNA. Such a system eventually came into existence—the
RNA viruses, whose genes are made of RNA and which parasitize the machinery of DNA
replication (the HIV virus is the most notable example of these viruses). Thus, the regularity
that all genes are made of DNA gives way to the regularity that they are all made of nucleic acids
(either RNA or DNA). But we can be sure that the arms race of evolutionary competition will
eventually undermine this new invariant regularity, by producing an alternative means of genetic transmission that exploits the regularity (unless it already has done so, by bringing about the
prion protein that transmits BSE, also known as ‘Mad Cow disease’). The same arms race
between DNA and RNA and prions also disposes of another invariance of molecular biology,
the so-called Central Dogma (in its strong form) that the flow of genetic information is always
from DNA to RNA to proteins.
Restricted Regularities in Social Sciences
13
can, natural selection will break up these local equilibria along with the regularities that describe them and their consequences.12
When traits are genetically coded, the arms race process will be relatively
slow, though much faster than non-arms race adaptational change. Since favorable mutations are rare, biological invariances between genetically
encoded traits will be often be locked in as the result of some relatively
long-term stable equilibrium in the arms race. Besides interspecific competition, there is also a great deal of intraspecific competition, arms races between
lineages within a single species, which also makes and breaks invariances at an
even faster rate than interspecific competition does.
The upshot is that all invariances among genetically encoded traits are restricted. During the periods that they obtain, they are vulnerable to being
undermined by random variations that break up co-adaptational equilibria.
As the rate of variation increases, the life span of an invariant regularity will
decrease as will the spatial range over which it obtains. In the case of competition between very fast breeding species—say, between parasites and their
host-targets, for example, phage and bacteria or bacteria and humans—
regularities may remain invariant only for a few years, months or even
weeks. (Consider the life-times of antibiotic effectiveness).
The role of the environment and of the onset of arms races in making and
breaking biological invariances explains clearly the differences in the successes
and limits of biological models. Among the most widely used models in biology is the Hardy–Weinberg model which is employed to describe gene and
genotypic frequencies of sexual species under a set of well-known assumptions, and can reliably be employed to identify sources of evolutionary change.
12
Natural selection even produces locally invariant regularities between traits that are not adapted
at all, but are correlated as the by-products of traits mutually selected for. For example, consider
a remarkable discovery of Darwin: in all mammalian species subject to domestication at least
some examples are ‘piebald’—i.e. have spots, usually white on dark—and this trait is heritable.
Darwin’s observation has since been widely confirmed: even ‘natural experiments’ in domestication of hitherto wild and non-piebald species such as the fox have produced this trait.
Presumably, being piebald is not an adaptation, and in general animal breeders do not select
for it. The relationship between being domesticated and being piebald is nevertheless invariant,
or has been hitherto. However, we pretty much know why. Domestication has always proceeded
by allowing the tamest, least aggressive young to reproduce with one another. Tameness is a
hereditary trait. At least some of the genes involved in tameness behavior are probably located
close together on the same chromosomes as recessive genes that control for variegated coat
color. Repeated interbreeding always brings out the recessive trait of piebald coat in at least
some descendants. So long as those chromosomes are not broken in meiosis at points between
the herd genes and the piebald genes, the regularity that domestic species have some piebald
members will be invariant. But of course this generalization is evolutionarily contingent. There
are several obvious circumstances—human and natural interventions—that can and some day
probably will break it down. Besides a suite of mutations, a founder effect recombination that
breaks the chromosomal link, or (equivalently) a persistent program of artificial selection to
breed non-piebald domestic animals, there is the possibility of a new move in some arms race we
have not noticed breaking the invariance. Being piebald may become an adaptational disadvantage owing to the conspicuousness or other fitness lowering effect of such marks in an
evolutionary arms race with predators or parasites.
14
Alex Rosenberg
There are only a small number of values of its variables for which the Hardy–
Weinberg model fails to be invariant. This is largely because intergenomic
arms race conflicts are few and far between, most having devolved into very
stable equilibria long ago on evolutionary time scales. A somewhat less invariant relation is described by Fischer’s model of why sex ratios remain equal
in sexually reproducing species. There are well-understood environments in
which the optimal sex ratios are far from equal and the model fails to apply—
for example, where the environment prevents female offspring from
dispersing. Relationships among predators and prey of the sort modeled by
Lotka–Volterra simultaneous differential equations are likely to be more
invariant than those between bacterial parasites and their hosts described in
Nicholson–Bailey models of a mathematically similar type. This is because in
the latter case there is a much shorter intergenerational time and more rapid
rate of mutation, which results in the parasite or host much more quickly
breaking out of the relationship between them altogether; in the former
case, where the generation time and the mutation rate are slower, predator/
prey relations are slower to break down.
In biology, restricted regularities are explanatory because they reflect local
equilibria underwritten by unrestricted regularities—laws of nature—that
Darwin discovered. These regularities are restricted owing to the inevitability
of arms races which the unrestricted regularities Darwin discovered ordain.
The next section explores whether appealing to Darwinian processes and
their consequent arms races can underwrite the explanatory power of restricted regularities in the social sciences.
5 Restricted Regularities, and Arms Races in Human Affairs
There is a fairly direct argument for the claim that restricted regularities in
human affairs are explanatory for exactly the same reasons that such regularities are explanatory in biology. The difficulty with this argument is grounding its premises. I provide the argument and sketch how its premises can be
grounded immediately afterward.
(1) Almost all significant features of human affairs—historical actions,
events, processes, norms, organizations, institutions, etc.—have
functions—i.e. adaptations, or else they are the direct results of
such adaptations.
(2) The only source of functions or adaptations in nature—including
human affairs—are Darwinian processes of blind variation and environmental filtration.
(3) All regularities among adaptations (or their direct results) are local
equilibria, which are eventually broken up by arms races.
Restricted Regularities in Social Sciences
15
(4) Such restricted regularities have limited explanatory power underwritten by unrestricted Darwinian regularities.
Therefore,
(5) All restricted regularities about human affairs are local equilibria
eventually to be broken up by arms races.
(6) Regularities about human affairs have explanatory power because
they are underwritten by unrestricted Darwinian regularities.
This argument provides the restricted regularities of the human sciences with
the explanatory power that such regularities have in biology. The cost of
adopting such an argument is a commitment to very strong Darwinism
about human affairs. The only premise that requires much argument
beyond what is already available in the philosophy of biology is the first
one. But can be given considerable support. Premise (1) may seem dubious
at first blush. How could almost everything in human affairs be an adaptation?
That sounds like an idea worthy of Pollyanna or Voltaire’s Dr Plangloss. Even
in biology, not everything turns out to be an adaptation. Much of evolution is
a matter of drift—the play of chance on small and sometimes even large populations that leads to changes in the distribution of adaptations, and even to the
persistence of non-adaptive and maladaptive traits. Moreover, important biological traits are themselves either the result of physical constraints or were
acquired as adaptations early enough in evolutionary processes to remain fixed
long after they ceased to be adaptations. Surely all the same must be said of the
course of human affairs. Indeed, for obvious reasons, there may well be a
greater role for drift and constraint in human affairs than biological processes.
Of course premise (1) does need to be understood as qualified by the reality
of drift and constraint in human affairs. In fact, the plausibility of the claim
that premise (1) makes about the adaptedness of most features of human
affairs relies a great deal on the qualification ‘significant’. There will be
many features of human affairs that are the result of drift, and yet few
social scientists will accept the suggestion that what particularly interests
them about human affairs is the result of random drift alone or even
mainly. Similarly, social scientists will recognize constraints of many kinds
as forcing subsequent features of human affairs to adapt to them. But few
social scientists accord such constraints the fixed character that constraints—
especially physical ones—have in biological evolution. In fact, the most revolutionary social changes in fact break down the oldest, firmest, and most
pervasive constraints, as a result of processes of variation and selection. The
real issue is whether such variation is blind and the resultant selection natural.
Reflection on human affairs does suggest that even more than in biology,
significant features of social life are largely or even wholly adaptations for
16
Alex Rosenberg
someone, or some group, or some practice. Human social life consists of
adaptations constructed by individuals and groups to cope with an environment that has mostly come to consist of other individuals and groups and their
adaptations. Some of the adaptations are ones people think they designed—
institutions like the US constitution, and built—artifacts like the Eifel Tower.
But mostly the adaptations emerged from history without any one intending,
designing, or even recognizing them. This is especially true of the most
important ones: think of feudalism, and the Roman Catholic Church, two
adaptations that were around for a long time. Then there are short-lived
adaptations—processes like the New Deal, occurrences like D-Day, or
events like the Defenestration of Prague, which are composed of parts someone did intend and ordain and other parts that just emerged as the unintended
consequences of intentional actions by some of their participants.
Then there are the features of human life that no one designed, that didn’t
emerge unintentionally from actions and events people did ‘design’ or intend,
but that are best thought of as symbionts, or parasites, or sometimes combinations of both, living on human life, and changing it for the better or for the
worse, but always adapting to insure their own survival.
It’s difficult to think of tobacco smoking, or heroin addiction as adaptations, because they are harmful. They are harmful to humans, but they are
practices with features that ensure their persistence and spread through
human history, at least until their environments change and their effects
start to be selected against. Chinese foot-binding is a clear example of how
this works. Foot-binding persisted for about 1000 years in China. It got
started because women with bound feet were more attractive as wives.
Bound feet were a signal of wealth, since only rich families could afford the
luxury of preventing daughters from working. Girls with bound feet were
easier to keep track of and so likelier to be virgins, etc. So, at first, when the
practice arose, foot-bound girls had more suitors. Pretty soon every family
that could afford it was binding daughters’ feet to assure they’d get married.
Result: when every girl’s feet were bound, foot-binding no longer provided an
advantage in the marriage market, and all foot-bound girls were worse off
because they couldn’t walk, suffered other health effects, etc. Foot binding
starts out as an adaptation for some girls, and for some families, but by the
time it becomes really widespread and fixed, it is actually a physical maladaptation that lowered every foot-bound girl’s fitness. But once every one was
doing it, no one could get off the foot-binding merry-go-round. Anyone who
stopped binding daughters’ feet condemned them to spinsterhood. Here we
have a tradition, a norm—bind daughters’ feet!—that by the time it was
widely adopted ceased to an adaptation for the people whose behavior it
governed. Why did it persist despite its maladaptive effects on foot-bound
girls? For whom or for what were its features adaptations? For itself, for
Restricted Regularities in Social Sciences
17
the practice, norm, institution of foot binding! The practice persists, like any
parasite, because of its adaptations, those of its features that exploit the
‘weaknesses’ of humans and their institutions—marriage, the desire for
virgin-brides and large dowries, the desire to control women before and
after marriage.13
Once we widen our focus, the claim that almost everything of interest to
social scientists in human affairs has functions or is an adaptation becomes far
less Panglossian. But can it be correct? One reason to suppose it must be is the
fact that almost all the vocabulary and taxonomy of common sense and the
human sciences are themselves thoroughly functional: their taxonomies individuate kinds in terms of their effects and (more rarely) their causes. Since
instances of these types persist over long periods, their effects are unlikely to
be harmful to them, still less lethal. Accordingly, social scientists have widely
assumed, the effects are beneficial to their persistence, they are their functions.
As a consequence social/behavioral sciences almost invariably taxonomize
their kinds in terms of their functions. Broad movements in social science
have erected this taxonomy into methodological practice: witness the traditions of functionalism, structural functionalism, and its attendant distinction
between manifest and latent functions. (See Rosenberg [2007], Chapter 7).
Moreover, the predictive and ameliorative goals of the human sciences
impose upon them all research programs that assume that most of the significant features of human affairs are adaptations for some individuals and
groups, and maladaptations for other groups. Though each of the social sciences may be neutral on the adaptive character of the actions, events, norms,
practices, and institutions in the domains of the other social and behavioral
sciences, it will not be agnostic about those within its proprietary domain. This
will be true at least so long as it has ameliorative ambitions for social processes
in its domain. The remodeling and redesign of political, legal, economic,
social, or cultural institutions, rules, norms, and practices would be impossible
if these items were not to varying degrees adaptive and maladaptive for individuals, and groups of various sizes and compositions, or for themselves as
parasites on human wants, needs, and interests.
To these considerations must be added the qualifications on premise (1) that
are required to accommodate drift and constraint. Just as these two factors
form a central part of contemporary Darwinian theory in biological science,
any extension of the theory to social science must be understood as including
drift and constraint as indispensible components of its explanatory resources.
If the scope for them, and especially for drift, is much greater than in biology,
then of course a Darwinian theory of human affairs will leave proportionately
13
See (Mackie [1996]) for an introduction to Chinese foot-binding in a game-theoretical framework. I am indebted here to Andres Luco.
18
Alex Rosenberg
more social phenomena as merely unexplained matters of random drift than
does a Darwinian theory of biological evolution. Great scope for drift would
certainly be a weakness of Darwinian social science. I doubt the scope for it is
in fact very great in those areas that interest social scientists.
Premise (2) asserts that only Darwinian processes can account for adaptations, whether biological or social. Here the argument is by exclusion of
alternative processes. There are obvious reasons to rule out several alternative
explanations for the ubiquity and persistence of adaptations in biological
and human affairs. Neither a designing deity nor future ends, goals, purposes,
or targets can be invoked in a causal explanation for any adaptation.
Immanent teleologies, entelechies, and vital forces are as objectionable as
future causation. This leaves only human design as an alternative to
Darwinian processes of blind variation and environmental filtration, as a
source of the adaptedness in human affairs.14
In general, people overestimate the role of intentional human design in
human affairs, and even this overestimate leaves the origin and persistence
of a large number of social adaptations unexplained, among which are some of
the most important ones. For example, consider Hayek’s ([1945]) Nobel-prize
winning identification of the function of the price mechanism and proof that it
could not be the result of any individual human intentions. Another example
is provided by Coase’s ([1937]) explanation of the origin and persistence of the
firm as an adaption for solving a transaction-cost problem that no one who
was party to the origin and maintenance of business firms probably ever
recognized.
Moreover, there are reasons to think that when human norms, practices,
and institutions are once consciously invented, their persistence can only be
the result of Darwinian processes, as opposed to continual intentional maintenance. And there is further reason to suppose that the processes of human
conscious intentional creation are themselves Darwinian ones carried out
within the brain(s) of those who intend them. For the explanation of the
apparently purposive creations of human intention faces the same problems
as the explanation of other biological adaptations: once we have ruled out
future causation, immanent teleologies, or vital/spiritual forces and disembodied minds, there seems no alternative to treating brain processes that eventuate in individual actions as Darwinian ones as well.
But if human affairs are mostly the emergence, persistence, and improvement of adaptations by Darwinian processes, then we can expect only a limited number of different kinds of regularities about the relationships between
14
Philosophers may add one more alternative source of apparent adaptation: Donald Davidson’s
‘swampman’ is sometimes invoked as a counter-example to selected effects, i.e. Darwinian accounts of adaptation. See (Davidson [1987]). The possibility this scenario invokes is not taken
seriously in social or behavioral science.
Restricted Regularities in Social Sciences
19
them, regularities of the sort already familiar in biology and exemplified in
Section 4 above. To begin with, in human affairs individual adaptations will
emerge, then there will be pairs or larger sets of adaptations each of which
constitute selective components of one another’s environment; they will either
accommodate to one another, or cooperate with one another, over long periods, or will compete with one another over such periods. In either of the first
two cases, any variations in either that can exploit to its selective advantage
the accommodation or the cooperation will do so. The result will eventually be
the emergence of a competition somewhere. This competition may persist in a
local equilibrium. But it must eventually break down into an arms race as
further variants of each adaptation emerge. The results of the emergence of
adaptations and of the persistence of these alternative relations between adaptations are the regularities we recognize in the social sciences.
First, there are regularities about the (real, or latent) function of a norm or
practice or institution: e.g. ‘all firms function to solve the transaction cost
problem’. Then there will be regularities about the co-occurrence of adaptations in the same individuals or lineages of them, regularities of co-occurrence
of adaptations in two or more distinct individuals or lineages—cases of
co-evolution or mutualism, or parasitism (I give the most famous example
of a regularity of co-evolution or mutualism in political science below). These
regularities will in effect record local equilibria among adaptations, ones that
last as long as the regularities that report them obtain.
But as in biology, each individual’s or group’s adaptation sets a design
problem for the individuals and groups with which it finds itself in local equilibrium. The existence of these mutual design problems together with the
persistent but blind variation among adaptations means that the prospects
for arms races are ever present and every local equilibrium must eventually be
broken by an arms race. Whence the restricted character of every explanatory
regularity and all the models in social science.
Examples of these are easy to identify. Consider perhaps the most robust
regularity in international relations, perhaps even in the whole of political
science: arguably, no two democracies have ever gone to war with one another. There is apparently not a single exception to this regularity since
democracies emerged at the end of the 18th century, even though the
number of pairs of countries at war with each other since 1776 is literally in
the several thousands.15 Apparently, the trait of being a democracy and the
15
See (Maoz and Russett [1993]). Rare dissenters from this view have invoked the US Civil War,
and the belligerency of Kaiser Wilhelm’s Germany as counter-examples (Layne [1994]), since
both parties to the civil war considered themselves democracies, and Germany had an elected (if
largely powerless) Reichstag. Suffice it to say that these counter-examples are controversial and
have been rejected on a variety of counts. If accepted they would not undermine the argument
advanced here.
20
Alex Rosenberg
trait of not going to war with a democracy are for the moment at least
co-adapted to each other, and this probably helps explain several things,
such as why democracies do better economically than even market-economy
dictatorships (fewer wars), and why the number of democracies seems to be
increasing. It has also guided foreign policy—the US and European encouragement of new democracies to ensure stability and peace.
But nothing is forever. We can be confident that somewhere or some when,
some democracy is going to find a way to exploit this regularity by attacking
some completely unsuspecting fellow-democracy, lulled into a false sense of
the permanence of peace among democracies. How can we be so confident?
Just as mother nature searches through biological design space for variations
that can take advantage of the environment of other adaptations the variation
faces, so too it will search for variations in the human social design space.
And the rate of variation will be vastly accelerated in comparison with
biological evolution. For, human evolution doesn’t have to wait a 20-year
generation for a variation in genes to change a trait the way that biological
evolution must.
One more example, this time of a mathematical model explanatory over a
restricted domain whose explanatory power is destroyed by an arms race
breaking up a local equilibrium. Consider again what was for a long time
the most influential model in economics, and perhaps even all of social science,
the IS/LM (Investment Saving/Liquidity preference Money supply) graphs
and equations of Keynesian macroeconomics. This set of graphs enabled
economists of the third quarter of the 20th century successfully to model
the invariant relationship between sets of macroeconomic variables, including
investment and savings, consumption and gross-national income, the interest
rate and the money supply. The Phillips curve discussed above, having been
grounded in macroeconomic data, was often explained by appeal to the
Keynesian macroeconomic model.
The stagflation of the 70s put an end to the model’s general acceptance, and
resulted in its replacement by newer ones, including the rational expectations
model. This model explained why the superseded model was no longer a basis
for effective intervention. The analysis of why the Keynesian model ceased to
work was roughly that, if it ever worked at all, then the co-adaptations it
identified were broken up by an arms race. The model’s widespread dissemination, or at least the fact that economic agents had become acquainted with
the governmental interventions it guided, resulted in a change in their choices,
one which rendered the Phillips curve inoperative and fiscal policy ineffective.
The beautiful and for a while powerful model of the capitalist economy that
John Maynard Keynes inspired ceased to work because the relationships
it described broke down once some of the institutions, groups, and individuals
the model included began to exploit the fact that other institutions
Restricted Regularities in Social Sciences
21
were guided by it, to frustrate the policies the model guided. Result: 10 or 15
years after it became widely known, the Keynesian model became the victim of
an arms race.
Examples are not arguments even though they can be proliferated ad infinitum. But there is a simple argument which they buttress and which explains
them. It shows exactly why restricted regularities and mathematical models
are explanatory and why they are the best we can hope for in the explanation
of human affairs.
6 Dealing with the No-memes Objection to
Darwinism about Human Affairs
The simple argument of Section 5 will be accused of invoking Darwinian
processes without showing that their necessary conditions obtain, and it will
be argued that a crucial one among these necessary conditions cannot obtain
in human affairs.
The objection is simple: Darwinian processes require replicators and interactors. In particular, it requires high-fidelity replicators to store and transmit
traits faithfully enough and long enough for environmental filtration to shape
them into adaptations and to maintain them as adaptations. In the biological
domain these replicators are the genes. There are no equivalent replicators in
human affairs, or at least there are not enough of them for Darwinian processes to explain human affairs. Darwinism of the sort evinced in the argument
of Section 5 requires cultural replicators—memes. But there are no memes.
No memes, ergo no Darwinian processes. (Cf. Sperber [2000] for the origin of
this widely mooted argument).
Darwinian social scientists can respond to this objection in many ways.
To begin with, it is obvious that the literal application of Darwinian theory
to a domain does not require that the domain contain gene-like replicators.
Biological replicators—genes—did not predate Darwinian processes. Natural
selection presumably got its start prior to the emergence of these replicators,
and for that matter, prior to the emergence of any recognizable very
high-fidelity informationally rich replicators. Moreover, natural selection is
likely to produce replicators of the sort that the genes constitute only when
environments change slowly, evolution is extremely gradual, cumulative, and
atomistic in its shaping of individual traits, one by one, for adaptations. When
one or more of these conditions do not obtain, adaptive evolution may employ
replicators and processes of replication quite unlike genes. (Cf. footnote
11 above).
Following Richerson and Boyd ([2006]), defenders of Darwinian
approaches to culture have also argued that cultural replicators need not
have the high-fidelity features of genes, since there are a variety of practices,
22
Alex Rosenberg
norms, and institutions in human culture which have emerged as adaptations
precisely because they preserve the adaptive informational content of replicators even under conditions of low copy fidelity. (Cf. Driscoll [2008] for a useful
recent discussion.)
It should however be granted that Darwinism about human affairs does
require cultural replicators, probably a variety of quite different kinds of
replicators, and some of them may be gene-like. I am prepared to call them
memes, since the label is such a good meme. But even if we grant that
Darwinism about human affairs requires at least some memes that are a lot
like genes, the arguments that there are no such things as memes are poor
ones. More important, the memes which Darwinism about human affairs
requires will be huge in number, short in life spans, and extremely difficult
to individuate, for the very same reason regularities in the human sciences are
restricted: because of the ubiquity of arms races. So, it will be no surprise that
few obvious examples of memes can be provided now, or perhaps ever.
To reprise the argument from no memes to no Darwinian processes in
human affairs; memes have to be like genes, because natural selection only
works in culture when there are gene-like replicators in culture. There are no
such gene-like replicators in human affairs. Ergo no Darwinian processes in
human affairs. The trouble with this argument is that it rests on an idea of
what genes are and how they work that was obsolete about a hundred years
ago. This is the one trait-one gene idea, the notion that most or many significant observed inherited traits are controlled by a single gene.
There are only a small number of such traits in any mammal, and in humans
only about seven such traits are known—for example, tongue rolling or the
widow’s peak. All other inherited traits in humans, like eye- and skin color,
even sexual characteristics, are each the result of the inheritance of many and
in some cases a huge number of genes. In fact, genes don’t really transmit or
control the appearance of any of the biological traits common sense and folk
biology think they do. Each gene controls the production of a protein or other
large molecule. There are 25,000 of these genes, switched on and off in every
cell of our bodies. It’s the protein molecules they code for, and the order in
which the proteins are produced by the genes that build biological machinery.
Many different traits that don’t have anything to do with each other are built
or controlled by the same gene; many traits that look absolutely the same to
us—say eye color—are the result of different sets of different genes in different
people. And when we actually locate the genes inside the nucleus of our
somatic cells, and in the sperm and egg that develop into our bodies when
they combine, the DNA structures of the same genes can differ from one
another without that difference making any difference to the proteins they
produce.
Restricted Regularities in Social Sciences
23
The moral for memes is obvious: if memes are like genes then any single
meme will by itself almost never control the appearance of a behavior or
action or anything else that common sense or even sophisticated social science
is interested in. It will take many, many memes working together to produce
anything of interest to the human sciences, and we will never be able to detect
or identify memes by doing anything like garden-variety sociology, economics, anthropology, or even psychology. Whatever memes are, they are going to
be as complicated and hard to identify as genes are, or even more so!
If memes are anything like genes, it is going to be very hard to identify them,
isolate or individuate them, and learn the details of how they work. Doing any
of these things for ‘memetics’ will be orders of magnitude more difficult than
what a century of molecular biology has done for genetics. And the reason for
all three of these difficulties will in large measure be the ubiquity of arm races
cutting short the life-time of any regularity about a particular package of
memes. The shorter this life-time, the more difficult it will be for any social
or behavioral scientist to design a method for identifying these packages, still
less identifying their component memes. To figure out the mechanism of transmission of some package of memes that controls for a socially interesting
adapted behavior will require first a long enough lived regularity about that
adapted behavior, and second, another long enough lived regularity about
how the particular package of memes that codes for the behavior and transmits it does each of these things. Since, in a social environment of accelerating
change, these packages of memes and the memes themselves will be transmitted faster and faster, and their modes of transmission and control interfered with more and more effectively, as arms races accelerate, the chances of
identifying and locating memes become lower and lower as cultural change
accelerates.
This is not an argument for memes. At most it is a well-grounded explanation for why social and behavioral scientists are unlikely to find them, an
excuse erected on the criticism of a bad argument against the very possibility
of memes. If there are memes, regularities about their transmission, mode of
action, and realizations in the brain will be complex, short lived, and completely beyond the reach of any hypothesis-testing in the social sciences.
So, if we can’t identify them, why suppose that there are any memes?
The real argument for memes is the two premises of Section 5. If these
premises are reasonable and support the conclusion that human affairs
must be Darwinian in nature, then that’s the argument for memes, or whatever
replicators are required by Darwinian processes in human affairs.
Should the human sciences cease work until neuroscience or some even
more molecular sub-discipline has established the existence or nonexistence
of memes? Did the rest of biology stop working while geneticists did the century of research that was required first to establish the existence of genes and
24
Alex Rosenberg
then to identify their location, composition, and model of action? Of course
not. The same goes for the human sciences.
7 Conclusion: A Moral for Institution Design
In one respect the arms race character of human interactions, together with
the decreasing cost of acquiring and employing information to make instrumentally rational decisions, may have a reflexive impact on social science,
mathematical modeling, and especially models that invoke rational choice
theory. In fact, the recent financial crisis and the consequent demand for
reforms of the credit markets provide a nice illustration of the reflexiveness
of social science and the arms races currently in progress in human affairs.
As we now know from the work of cognitive psychologists, evolutionary
game theories, and others, human choice-behavior is mostly the result of
heuristics, fast and frugal inference rules, preferences that do not honor von
Neumann–Morgenstern axioms of expected utility, and norms of fairness,
equity, and cooperation that emerged as a result of natural selection in
pre-modern settings.
In any competition with users of rational choice models, the losses due to
employing quick and dirty heuristics in decision making rise, and the incentives to engage in classically rational behavior increase. As the incentives to
employing rational choice models increase and the costs of doing so decline,
more and more of the behavior of interest in the social sciences—political
behavior, business strategy, social cooperation—will begin to satisfy the
models of rational choice theory.
This will put pressure on many cooperative institutions that emerged in past
environments as evolutionarily stable Nash equilibria among individuals
making choices employing a variety of heuristics, including fairness, equity,
and other cooperative norms. Under relentless attack from rational choice
norms, such cooperative institutions will have to undergo substantial redesign.
Otherwise they will fall prey to arms race subversion, and fail to protect participants from perfectly rational free riders operating in the new environment
created by the changed costs and incentives. Institutions that don’t change to
protect their participants will be undermined and will become extinct, along
with the regularities in behavior that they support. Institutions that do change,
and altogether new institutions, embodying new norms, may persist in new
equilibria for varying lengths of time. But eventually they will simply set new
design problems for participants and others with incentives to exploit them,
and thereby set off a new sequence of arms races.
If substantial financial reform can be introduced, the costs of coming up
with good new ideas about how to beat the system will increase or the benefits
Restricted Regularities in Social Sciences
25
of doing so decrease enough to slow down this arms race that nature imposes
on all its creatures. But don’t count on its happening any time soon.
Acknowledgements
Thanks to James Woodward, Philip Kitcher, Kevin Hoover, Elliot Sober,
Ken Waters, Carl Hoefer, Kim Sterelny, Don Ross, and an anonymous
referee for helpful comments on previous drafts.
Department of Philosophy
Center for Philosophy of Biology
201 West Duke Building
Duke University Durham
NC 27708, USA
[email protected]
References
Beatty, J. [1995]: ‘The Evolutionary Contingency Thesis’, in G. Wolters and J. Lennox
(eds), Concepts, Theories, and Rationality in the Biological Sciences, Pittsburgh:
University of Pittsburgh Press, pp. 45–82.
Coase, R. H. [1937]: ‘The Nature of the Firm’, Economica, 4, pp. 386–405.
Davidson, D. [1987]: ‘Knowing One’s Own Mind’, in Proceedings and Addresses of the
American Philosophical Association, 60, pp. 441–58; reprinted in his Subjective,
Intersubjective, Objective, New York: Clarendon Press, pp. 15–38.
Dixit, A. and Skeath, S. [2004]: Games of Strategy, New York: W.W. Norton &
Company.
Driscoll, C. [2008]: ‘The Problem of Adaptive Individual Choice in Cultural Evolution’,
Biology and Philosophy, 23, pp. 101–13.
Fodor, J. [1974]: ‘Special Sciences and the Disunity of Science as a Working
Hypothesis’, Synthese, 28, pp. 77–115.
Friedman, M. [1968]: ‘The Role of Monetary Policy’, American Economic Review, 58,
pp. 1–17.
Hayek, F. [1945]: ‘The Uses of Knowledge in Society’, American Economic Review, 35,
pp. 519–30.
James, J. [1989]: ‘The Stability of the 19th Century Phillips Curve Relationship’,
Explorations in Economic History, 26, pp. 117–34.
Kahneman, D., Slovic, P. and Tversky, A. [1982]: Judgment Under Uncertainty:
Heuristics and Biases, New York: Cambridge University Press.
Keynes, J. M. [1936]: The General Theory of Employment, Interest and Money, London:
Macmillan.
Kitcher, P. [1993]: The Advancement of Science, New York: Oxford University Press.
Layne, C. [1994]: ‘Kant or Cant: The Myth of the Democratic Peace’, International
Security, 19, pp. 5–49.
26
Alex Rosenberg
Lucas, R. [1972]: ‘Expectations and the Neutrality of Money’, Journal of Economic
Theory, 4, pp. 103–24.
Mackie, G. [1996]: ‘Ending Footbinding and Infibulation: A Convention Account’,
American Sociological Review, 61, pp. 999–1017.
Maoz, Z. and Russett, B. [1993]: ‘Normative and Structural Causes of Democratic
Peace, 1946-1986’, American Political Science Review, 87, pp. 624–38.
Phillips, A. W. [1958]: ‘The Relation Between Unemployment and the Rate of Change
of Money Wage Rates in the United Kingdom, 1861–1957’, Economica, 25,
pp. 283–99.
Richerson, P. J. and Boyd, R. [2006]: Not By Genes Alone, Chicago: University of
Chicago Press.
Rosenberg, A. [2007]: Philosophy of Social Science, Boulder: Westview/Harper Collins.
Sperber, D. [2000]: ‘An Objection to the Memetic Approach to Culture’, in R. Aunger
(ed.), Darwinizing Culture: The Status of Memetics as a Science, New York: Oxford
University Press, pp. 163–73.
Woodward, J. [2000]: ‘Explanation and Invariance in the Special Sciences’,
British Journal for the Philosophy of Science, 51, pp. 197–254.
Woodward, J. [2003]: Making Things Happen: A Theory of Causal Explanation, Oxford:
Oxford University Press.