Greg Detre
Monday, 5th November, 2001
Nagel defines rationality as the capacity �to recognise objectively valid reasons and arguments�. He is a realist about reason and does not want to accept reductive interpretations of it is as a �contingent though basic feature of a particular culture or form of life�. He recognises that he needs to address the question of how it is that contingent, biological creatures such as ourselves can have access to such universally valid methods of objective thought. I want to focus on the extent to which evolutionary theory supports or undermines such rational objectivism, as I will term Nagel�s position.
The fundamental tenet of evolutionary theory (or �neo-Darwinism�) is the principle of natural selection, whereby parental characteristics that vary across organisms play a role in non-random differential reproduction. That is, �adaptive� variations are those which increase an organism�s (or other organisms with similar genes, strictly speaking) propensity to survive and reproduce (its �fitness�). This process gradually gives rise to diverse forms leading ultimately, through selective adaptation to specific niches and environments, to the emergence of new species. If humans are the end-product of a natural, non-teleological process of evolution that has resulted in the particular, contingent bodies and brains that we have, then our reasoning abilities and limitations will be explicable in terms of evolutionary theory too.
Curiously, evolutionary theory has been employed by both objectivists and subjectivists to support their claims. For example, Stich cites Quine, Dennett and Fodor as implying that evolution selects for rationality and that irrationality is empirically impossible or unlikely, e.g.:
creatures inveterately wrong in their inductions have a pathetic but praiseworthy tendency to die out before reproducing their kind.[1]
However, it seems more plausible that the opposite is the
case, i.e. that evolved creatures (including ourselves) are highly unlikely to
be objectively rational. I will start by reviewing how objectivists try and use
evolutionary theory to directly support their claim, and then try to show that
it actually heavily undermines their case. Accepting that the empirical
psychological and biological evidence clearly undermines the objectivists,
Nagel argues instead that such evidence belongs to the family of arguments that
can be attacked as self-defeating. Unfortunately, if
we accept his argument, he leaves us in a position analogous to the
�explanatory gap� in philosophy of mind, where the current explanations and
paradigms of hard science are irreconcilable with how things clearly seem to
the objectivists to be.
Stich considers that there are two main reasons why people think that evolution insures rationality:
He systematically takes apart both of these premises. His aim is not to demonstrate conclusively that evolution theory is incompatible with rational objectivism, but merely to show that the rational objectivists have their work cut out in showing how the two can sit together. I will not consider the arguments in as much detail, but I will try and recount, evaluate and supplement them to a limited extent.
The notions of �fitness� and �optimality� are central to any evolutionary theory. A system is �well-designed� if it enhances fitness more than any alternative. Of course, this is problematic because of the difficulties of deciding what counts as an alternative. No doubt a predator which had evolved a high velocity rifle as an extra limb would be at an enormous evolutionary advantage, but as I will discuss below, the march of evolution is restricted to a sequence of gradual changes, each adaptive in their own right � although the end result of a fully-functional rifle would be highly adaptive, all of the intermediary stages (growing a long, perfectly straight, protruding barrel, the combustion mechanism, an organ for ammunition manufacture etc.) would be highly maladaptive right up until they were all brought together.
Stich considers a number of technical arguments against a na� faith in evolution as an infallible optimiser. He points out that natural selection is not the only process that causes changes of gene frequency in populations (which is how biologists define evolution). Mutation, migration and random drift all affect gene frequencies, to a greater or lesser degree. For instance, a random event or disaster might wipe out a large proportion of a population, including all the carriers of a particular fit gene, allowing a less fit gene to take hold in the population.
Natural selection does not necessarily choose the best genes in the gene pool anyway � Stich discusses meiotic drive, the effect of combined recessive genes, pleiotropy and heterozygote superiority. Each of these phenomena can lead to less optimal members of the available gene pool being selected for. In the case of meiotic drive, for example, �certain genes have the capacity to �cheat� in meiosis [�the process that produces sperm and eggs�] and end up significantly over-represented in the sperm or eggs�, and so �obviously, such a gene will spread quickly through a population, even if the phenotypic effects of the gene are harmful�.
I would like to emphasise a deeper problem about evolution for objectivists that Stich touches on. Evolution is not teleological � it does not modity towards an end-goal, but by producing various slightly different incarnations, each one a little bit more or a little bit less adaptive than its parents. Most modern biologists are gradualists, that is, they believe that a host of these tiny changes amount to large-scale improvements in the organisms, like a new organ. In contrast, saltatory explanation holds that sudden, marked �leaps� (macro-mutations) drive evolution forwards. Assuming the gradualists to be correct, evolution suffers from what artificial intelligence researchers term the �hill-climbing problem�. In outline, this considers that the space of possibilities which evolution is exploring through natural selection working on population variation can be visualised as a landscape, in which the hills and mountains represent more optimal designs. If the mechanism of evolution progresses incrementally, visualisable as travelling in a continuous path uphill, then a population may become stranded on a nearby hill, unable to traverse to the nearby mountain (an optimal design) which is separated by a valley.
Nagel rightly states that �the essential characteristic of reasoning is its generality�.[2] Nagel is certainly not content to regard his rational capacity as forming beliefs that are merely relative, true for a New Yorker, true for an American, true for a human even. He seeks to defend reason as a universal source of authority. In order to do that, surely our inferential system needs to have evolved to be absolutely optimal. If it was merely very well-designed, it might be able to generalise well, but not generalise to �objectively valid� conclusions. Its conclusions might be much less subjective than a less well-designed system, but still not entirely generalisable. Given how unlikely it is that any evolved system is ever wholly optimal, does this mean that we there is no hope that we could be rational in the way that Nagel envisages? I think that the answer to this is probably �no�. In the discussion above, Cherniak showed that there is a scale of rationality, and that Nagel�s requirements for objective rationality place him at the high end, but not the top of that scale, requiring a high level, but not maximum, optimality.
Lastly, it is necessary to show that our cognitive system is a product of biological evolution. After all, �even if it were the case that natural selection is a flawless optimiser and that it is the only cause of biological evolution, it would still not follow that our system of inferential strategies is optimally well-designed� unless �evolutionary factors are the only [or major] ones that have shaped our current inferential strategies�. In order for natural selection to shape a characteristic, there must be variation in the population that affects reproductive success in a systematic way, and this variation must be under genetic control either directly or indirectly. He considers clothing styles and language as examples in which there is great diversity within the human population, that may have some impact on fitness, but this diversity is not genetically based. �Had I been born elsewhere, I would now have the ability to speak Lapp or Korean rather than English.� And the processes by which languages spread are almost entirely independent of biological evolution, depending far more on social and historical factors, for example. Similarly, it may be that �the [inferential] strategies a person employs, like the language he/she speaks, are determined in large measure by environmental variables�.
Assuming that the objectivist could show that evolution produces near-optimal systems, the crucial second step necessary to show that evolution selects for rational systems is to show that an optimal system is a rational one.
The first way of doing this uses the traditional methods of analytic epistemology. Analytic epistemology is the approach of grounding cognitive assessment in the analysis or explication of our ordinary evaluative concepts. It involves showing that �optimality� and �rationality� are conceptually, analytically identical, that is, �if we analyse what we ordinarily mean when we say that one inferential system is more rational than another � we will find we mean that one is more fitness enhancing than the other�, i.e. �the claim that optimally well-designed cognitive systems are rational is a conceptual truth�. Stich wants to bring the entire edifice of analytic epistemology crashing down, so naturally he has little time for this approach. Moreover, it certainly won�t satisfy Nagel�s demands unless the next argument, that having true beliefs is adaptive, holds as well.
eh???
The second means of showing that an optimal system is a rational one is to argue, in the same vein as the reliabilist accounts of epistemic justification, that �the rationality of an inferential system is a function of how well it does at producing true beliefs� (and avoiding falsehoods). In this case, we need to show that inferential strategies that generally yield true beliefs are fitness enhancing, i.e. having true beliefs is more adaptive. Of course, there will always be cases when having a particular true belief is not fitness-enhancing, e.g. getting the time of your train wrong, arriving late and missing it, only to find that it crashed en route, in which case you would have died.
true vs objective beliefs � both
reliabilist???
However, subjectivists need to show that a generally less reliable system could exceed a more reliable system. Following Stich, I will utilise Sober�s distinction between internal and external fitness.
The internal fitness of an inferential system relates to how economically it achieves its effects, in terms of the demands made on the organism�s memory, processing capacity, energy/resources, time etc. The declining marginal utility of increasingly accurate beliefs has to be considered when evaluating an expensive, reliable system as opposed to a less expensive, less reliable system. A good example of such a trade-off might be the fact that our cognitive representation of space employs Euclidean geometry, which isn�t as �true� as Einsteinian relativistic geometry, in that the latter provides a better quantitative model of the world. It seems reasonable to think that a relativistic representation of space would require considerably more processing capacity, in that it would require the agent to take account of the speed of an object when predicting how heavy it will be to pick up, for instance. The deviations between a relativistic and a Euclidean system are quite negligible at the human scale, and would offer almost no selection advantage whatsoever. Clearly, we would be highly likely to evolve towards any such cognitively cheaper, equally adaptive though less true system.
The external fitness of a system relates to how conducive to survival and reproductive success it is. One might consider the greater adaptive value of an over-sensitive strategy which produces plenty of false positives which have only minor negative implications for fitness but assiduously avoids potentially fatal false negatives. Such a system might well yield false beliefs more and true beliefs less, but still be favoured by natural selection, preferring �reliability-when-it-counts-most� over �overall-reliability�.
Nozick�s account supplements Stich�s arguments nicely. He sees the relationship between evolution and rationality as informing problems aired by Descartes and Hume. Hume�s problem of induction addressed the impossibility of finding a rational (deductive) argument for why (inductive) reasoning works. Descartes questioned why self-evident propositions, as discerned by the natural light of reason, must correspond to reality.
how are Descartes and Hume related
here???
After all, if reason and the facts are independent variables, why should they be correlated at all? Kant responded that since we cannot show why our reason would conform to objects, it must be that we perceive the objects to be the way they are because they are constructed by our faculties. In other words, our knowledge is not of things in themselves, but only of an empirical reality shaped by our constitution. Kant termed this upheaval a �Copernican revolution�, although confusingly, in contrast to Copernicus� effect on astronomy, Kant is reaffirming an anthropomorphic perspective on the universe.
Nozick is overturning the Kantian dependence of the facts on reason. His �evolutionary hypothesis� amounts to saying that it is reason that is the dependent variable, shaped by the facts. Our inferential system has evolved to become specialised for common past situations and stabilities in our environment. He suggests that there was selection for recognising as valid certain kinds of connections that are factual, which come to seem to us as more than just factual. Thus, the neural architecture for a given factual connection that appears regularly and stably in our environment may be modified over evolutionary time so that our descendants learn it faster.
Thus, �reason tells us about reality, because reality shapes reason, selecting for what seems �evident��. However, just because a certain factual connection has been consistent in the past and we have evolved to see it as a vaild and increasingly self-evident basis for inference, does not guarantee that it will continue to hold in the future. Moreover, the question is not just whether the stable regularities of the past continue to hold in the future, but also whether evolution has picked out the �right� regularities or given us �green� in a �grue� world. We may come to see the given sequence of thought as increasingly self-evidently certain (because we are selected to do so, because in a stable world such semi-automatic inference-making is adaptive), but this does not guarantee that it ever was strictly true.
Importantly then, he is not saying that it is the capacity to recognise independently existing valid rational connections that is selected for. Rationality can be seen as a biological adaptation with a function. It was never the function of rationality to justify certain of our most basic, stable, useful assumptions, because all we needed was to utilise them as trust-worthy, predictive regularities. These basic, sub-rational assumptions include the list of philosophical problems we�ve been least successful with: the problems of induction; of other minds; of the external world; and of justifying rationality. It initially seems as though this wholly undermines any claims to objectivity through reason, though in fact Nozick is keen to defend against seeing this as the consequence of his arguments.
We may still be able to sharpen our goals and procedures though, at least to some extent. Evolutionary theorising may help us understand what sort of rational system would be adaptive, and consequently why our rational system is the way it is. The fact that rationality wasn�t designed to justify itself or its framework assumptions does not mean that it can�t, or that we can�t turn our rationality upon itself. Rationality is self-conscious, in that it attempts to correct biases in the information it is supplied, and in its processes of reasoning. Nozick claims that �Whatever the initial functions of reasons were, we can use our ability to employ reasons to formulate new properties of reasons and to shape our utilisation of reasons to exhibit these properties. According to Nozick, we can, that is, modify and alter the functions of reasons, and hence of rationality.� After all, although psychological experiments show how often most people fail to reason well, for example about probability, the very fact that we have been able (through centuries of reflection) to formalise and so correct such faulty reasoning lends hope to improving upon these biologically-instilled assumptions, e.g. Euclidean geometry.
However, Nozick stresses that his evolutionary account of why we find certain thought processes rationally self-evident does not provide a reason-independent explanation of reason, since after all, �the evolutionary explanation itself is something we arrive at, in part, by the use of reason to support evolutionary theory in general and also this particular application of it � Hence, the account is not part of first philosophy; it is part of our current ongoing scientific view.�[3] By conceding a lower epistemic status for the account, he avoids any circularity in his explanation.
However, although Nozick regards his account as a �proposal of a possible naturalistic explanation of the existence of reason that would, if it were true, make our reliance on reason �objectively� reasonable�, I think that Nagel is right to feel that �the idea that our rational capacity was the product of natural selection would render reasoning far less trustworthy than Nozick suggests�. As he says, in order for his objectivist position to hold, �I have to be able to believe that the evolutionary explanation is consistent with the proposition that I follow the rules of logic because they are correct � not merely because I am biologically programmed to do so�.
Nagel�s response to this and every evolutionary account is to regard any subjectivising evolutionary explanation of reason as a priori self-defeating. He follows Plantinga in saying that �it is irrational to accept evolutionary naturalism, because if it were true, we would have no reason to rely on the methods by which we arrive at it or any other scientific theory�.[4] �The recognition of logical arguments as independently valid is a precondition of the acceptability of an evolutionary story about the source of that recognition�. As a result, he concludes that any evolutionary explanation of rationality is necessarily incomplete. Even an account that is compatible with his objectivism can play no epistemic role in a justification of reason, unless it can be grounded itself on a basis other than that which it is seeking to prove.
But this leaves us in an unsatisfactory limbo. Nagel is prepared to admit that he doesn�t have a proper positive response to the question of how we can �integrate such an attitude towards reason with the fact that we are members of a biological species whose evolution has been shaped by the contingencies of natural selection�. He seems to place his hope in the thought that �it is possible that rationality � the capacity to recognise objectively valid reasons and arguments � is a distinctively accessible member of the set of biological possibilities, one that becomes likely at sufficiently high levels of biological complexity�. He understates the truth when he says that �the theory of evolution as usually understood provides absolutely no support for this conception of ourselves [�as rational � and also as creatures who have been produced through Darwinian evolution�], and to some extent it renders the conception suspect.�
An evolutionary argument which did make an objectivist account of rationality less mysterious would be reassuring, but still would not �underwrite� our use of reason in a reason-independent way. Ultimately, according to Nagel, �naturalistic accounts of reasoning� seek an external �understanding of the world [which] could close over itself by including us and our methods of thought and understanding within its scope�, but �this hope cannot be realised, because the primary position will always be occupied by our employment of reason and understanding � even when we make reasoning the object of our investigation�.
Interestingly, coming from a very different angle, the structure of Pinker�s theory about the evolutionary origins of our reasoning capacities dovetails nicely with Nagel�s speculative requirements for an objectivist-compatible evolutionary account:
However, it is ambiguous whether Pinker�s theory supports a fully objectivist picture of rationality, because he is more interested in how our natural scientific propensities differ from the academic approach. His aim is to show how an evolutionary account can explain modern empirical experiments showing the limitations and successes of people�s reasoning, as well as considering the original adaptive role of the faculties we can use for science, maths, chess etc. He argues that we �[build] parochial inference models that exploit eons-old regularities in their own subject matters�, e.g. �recognising objects, making tools, learning the local language, finding a mate, predicting an animal�s movement, finding [our] way� � the �subject-specific intelligence of our species� that Tooby and Cosmides[5] call �ecological rationality�. He argues in terms of both internal and external fitness (though not in those terms) that our brains have been shaped for fitness, not for truth.
Following Jackendoff, he considers these sentences (amongst many others):
The messenger went from Paris to Istanbul
The inheritance finally went to Fred
The light went from green to red
The meeting went from 3:00 to 4:00
Pinker argues that our entire linguistic faculty is based on concrete inferential machinery that has been co-opted to represent new, more abstract domains. The concepts of space and force �appear to be the vocabulary and syntax of mentalese, the [combinatorial] language of thought�. He speculates whether if �ancestral circuits for reasoning about space and force were copied, the copy�s connections to the eyes and muscles were severed and references to the physical world were bleached out�, �the circuits could serve as a scaffolding whose slots are filled with symbols for more abstract concerns like states, possessions, ideas and desires�. As evidence, he considers Premack�s experiments on chimps which showed that they could pick out the object which plays a causal role linking before-and-after pictures. Also, �space and force metaphors have been reinvented time and again in dozens of language families across the globe�. �Preschool children spontaneously coin their own metaphors in which space and motion symbolise possession, communication, time and causation� (Bowerman), e.g. �Can I have any reading behind the dinner?�. Thus, our minds aren�t really adapted to think about arbitrary abstract entities, so much as having inherited �a pad of forms that capture the key features of encounters among objects and forces, and the features of other consequential themes of the human condition such as fighting, food and health�. We can adapt these inherited forms to more abstruse domains.
In a similar way, mathematician Sanders Mac Lane followed Nagel�s two-part prescription of breaking rationality up into functional elements and considering what selective advantages these modules might have had. Mac Lane �speculated that basic human activities were the inspiration for every branch of mathematics�:
counting���� ��� arithmetic
measuring� ��� real numbers, calculus, analysis
shaping ����� ��� geometry, topology
forming������ ��� (as in architecture) symmetry, group theory
estimating� ��� probability, measure theory, statistics
moving������ ��� mechanics, calculus, dynamics
calculating ��� algebra, numerical analysis
proving������ ��� logic
puzzling����� ��� combinatorics, number theory
grouping���� ��� set theory, combinatorics
But we may not be biologically designed for (and it would be surprising if we were) large number words, large sets, the base-10 system, fractions, multicolumn addition, carrying, multiplication/division, radicals and exponents. These skills develop slowly and unevenly, perhaps by applying the sense of number to things that at first feel like the wrong kind of subject matter, and by practicing (chunking and automaticity � fitting together over-learned routines).
If anything, Pinker�s success at expounding a probably non-objectivist account while more or less adhering to Nagel�s prescription highlights the difficulty for an objectivist of explaining how objective rationality would/could conceivably have evolved. Indeed, Nagel�s position at the end of The Last Word seems somewhat analogous to his �explanatory gap� position in philosophy of mind, where current naturalistic accounts are insufficient to ground rationality in the way that it requires, i.e. �simply in itself � that is, believing what it tells me, in virtue of the content of the arguments it delivers�. Unfortunately though, I find his objectivist instincts about reason considerably less defensible and intuitively evident. In this case, however, he does have the luxury of being able to strike down any subjectivising accounts as a priori self-defeating, no matter how much more plausible, predictive, explanatorily powerful and integrated with the rest of our world picture such accounts may be.
�I don�t have [an alternative], and
I don�t need one in order to reject all existing proposals as improbable�.