:: Article

Darwinian conundrums

Interview by Richard Marshall.


Jonathan Birch is a brooder on the philosophy of biology. Here he thinks about teleological language use in science, about what natural selection can do, about Bill Hamilton’s kin selection theory, on what was in Darwin’s original theory and what wasn’t, about the origins of human cooperation, about the role of philosophy in science and about Nick Bostrom’s simulation argument. Step out into his jungle…

3:AM: What made you become a philosopher?

Jonathan Birch: A hatred of lab work! My first degree was in Natural Sciences at Cambridge. It’s a wonderful degree for the curious, because you can mix and match courses from all over the sciences. One of the courses I took was in the History & Philosophy of Science. And what a course it was: Peter Lipton taught the Philosophy of Science half, and made it sound like the most important and interesting subject on Earth.

I came to realize that the questions I cared about most were questions that only philosophy could answer, because practising scientists set them aside as being too foundational or too conceptual to grapple with. More importantly, though, I enjoyed writing and hated working in a lab. Or anywhere else, for that matter. I still enjoy writing, and I don’t really see it as a form of work.

3:AM: You work in the philosophy of science and in particular have focused on issues arising in biological sciences. So perhaps we should start by looking at a basic concern you’ve investigated, the issue of teleological language in science. One of the ways science is often distinguished from other ways of trying to explain reality is that it doesn’t use such language. Yet we do find it in science nevertheless. So why does teleological language infect the sciences? Is it just loose talk, shorthand that is always reducible to non-teleological talk, or is there a problem?

JB: I think ‘teleological’ language (i.e. talk of goals, purposes, and so on) is a shorthand, yes. A shorthand that seems natural—maybe even irresistible—for certain types of system. Organisms are the most obvious examples. Even very simple organisms with no brains—such as bacteria or amoebae—often look like they are pursuing goals, in the sense that they converge robustly on particular end-points (e.g. prey) despite encountering obstacles along the way. Talk of goals seems natural here, but there’s nothing spooky going on. One way or another, it’s all the product of evolved mechanisms.

What really interests me are more surprising, harder-to-explain examples of ‘goal talk’ in science. For example, biologists often seem to talk about populations in goal-directed terms. To take an actual example, a biologist might say that ‘ants evolved self-sacrifice in order to minimise the risk of disease transmission’. There’s no individual ant that evolved self-sacrifice: a population evolved it.

At face value, it sounds crazy to talk of populations pursuing goals. And yet I think it does make sense, as a shorthand, if you have in the back of your mind a picture of evolution in which populations are robustly converging on the peaks of an ‘adaptive landscape’. If the convergence is robust enough, then it will look like it is goal-directed from a sufficiently zoomed-out vantage point.

So I think what the prevalence of such talk implicitly reveals is the pervasiveness of this kind of ‘adaptive landscape’ thinking. One sees exactly the same language used in chemistry and biochemistry, when scientists talk about the robust convergence of a molecule on the minimum of a potential energy surface.

3:AM: You say that such shorthand is more applicable in the physical sciences than the biological – does this mean the biologists have no excuse?

JB: This was a claim I made in a paper on this issue (‘Robust processes and teleological language’) in the hope of stirring up controversy. I was referring to the kind of ‘goal talk’ I described above, in which it’s populations rather than individuals that are allegedly pursuing goals. I think this is dubious in biology, because I think the picture of evolution on which it tacitly relies (the ‘adaptive landscape’ picture) is a problematic one. In chemistry and biochemistry I have no problem with it, because I have no problem with the concept of a potential energy surface.

3:AM: Natural selection is a process that evolutionary biology uses to explain the traits of individual organisms. Some have doubts though: Elliott Sober, for example, argued that natural selection couldn’t do that. Can you first sketch what he argued?

JB: This is quite a subtle issue. We need to distinguish two types of question here. Let’s call them ‘type-1’ and ‘type-2’ questions:

Type-1: Why does a certain trait (e.g. red hair) have a certain frequency in a population?
Type-2: Why does an individual organism have one trait (e.g. red hair) rather than an alternative (e.g. brown hair)?

Sober’s view, often called the ‘negative view’ of natural selection, is that natural selection helps us answer type-1 questions but is irrelevant to type-2 questions. Only facts about development, inheritance and mutation matter to a type-2 question.

The reasoning is roughly like this. Take a trait of an individual organism, e.g. my red hair. Now ask: why is my hair red rather than brown? The thought is that, for natural selection to matter here, it would have to be that, if the selection pressures acting on my ancestors had been different, then my hair colour would have been different. But natural selection just doesn’t work like that. Altering a selection pressure can cause some individuals to die and others to live, some to reproduce less and others more, but it can never cause (well, except under unusual circumstances, which I won’t go into) the very same individuals to develop different traits.

3:AM: But you disagree with this, don’t you? Why?

JB: My view is that natural selection is sometimes relevant to these type-2 questions. To explain why, I need to say a bit more about the picture of causal explanation that’s in the background.

On the picture of causal explanation I favour, to ask why an organism has some trait rather than a specified alternative is in effect to ask: Under what biologically plausible conditions could the same organism have come to possess the alternative trait? And how does reality differ from that ‘counterfactual’ scenario?

When we answer these questions, natural selection isn’t always relevant, but sometimes it is. Crucially, for natural selection to be relevant on this view, it doesn’t need to be the case that we can alter an individual’s traits simply by manipulating selection pressures.


For example, there’s a type of mantis in South East Asia (Hymenopus coronatus) that looks exactly like an orchid. Suppose we go to the rainforest, single out a particular individual (call it mantis M) and ask: Why does M resemble an orchid rather than a primrose? To answer this, we need to construct a biologically plausible model of the circumstances in which M might have come to possess primrose-like camouflage, and then we need to see how that model differs from reality.

First, it would have to be a model in which a number of actual mutation events in M’s ancestral lineage—events that cumulatively led to its actual, orchid-like camouflage—never occurred. Instead, a number of counterfactual mutation events—gradually producing primrose-like camouflage—would have been needed.

Second, it would have to be a model in which the selective environment was different. The actual selective environment experienced by M’s ancestors, being full of orchids and devoid of primroses, favoured orchid-like camouflage. A lineage of mantises with increasingly primrose-like camouflage, leading ultimately to M, could not plausibly have persisted in that environment. So in order to construct a plausible counterfactual model in which a primrose-like M comes into existence, we need to alter the selective environment experienced by M’s ancestors to replace orchids with primroses.

This slightly wacky example is intended to illustrate a broader point: biologists think natural selection matters when explaining the adaptedness of an individual organism to its environment, and I think they’re right about this. I entirely agree with Sober that one cannot (normally) manipulate an individual’s traits simply by manipulating selection pressures. But, as I see it, this is not the only way for natural selection to be relevant to the explanation.

3:AM: Bill Hamilton’s theory of kin selection is a key idea in evolutionary biology. It’s had controversies. First could you sketch what the theory claims.

JB: It’s a lovely idea. The basic insight is this: when interacting organisms share genes, they sometimes have an evolutionary incentive to help each other. Crucially, the size of the incentive to help is proportional to the degree of relatedness between them. It’s an insight captured nicely in a quip by J.B.S. Haldane, who remarked that he would lay down his own life “for two brothers or eight cousins.”

The quip is Haldane’s, and the term ‘kin selection’ was coined by John Maynard Smith, but the theory itself is all Hamilton’s. Hamilton showed that you can derive a simple mathematical condition, now known as ‘Hamilton’s rule’, that specifies the circumstances in which a social trait will be favoured by natural selection. The condition is rb > c, where ‘c’ is the fitness cost to the organism that has the trait, ‘b’ is the fitness benefit the trait confers on another organism, and ‘r’ is the genetic relatedness between the two organisms.

The main qualitative prediction of the theory is that, when we find an organism performing a costly helping behaviour, we should expect to find that the benefit falls on its genetic relatives rather than on genetically unrelated organisms. This is indeed what we find. In social insects like ants and termites, in bacteria, in amoebae, in social mammals like wolves, chimps, gorillas, baboons, meerkats… and even, to some extent, in humans.

3:AM: So it’s been under fire from various people, including M.A. Nowak, C.E. Tarnita and E.O. Wilson the ant man. So why do they think it’s a rule that almost never holds? That’s a pretty fatal flaw isn’t it?

JB: The key is to get clear about what we mean by Hamilton’s rule. There are various different versions, and on some versions the rule almost never holds, whereas on other versions it almost always holds. Nowak et al. claimed it almost never holds because they had one of the more fragile versions in mind. My paper ‘Hamilton’s rule and its discontents’ sets out some of the mathematical details.

This debate, which may initially seem rather abstruse and mathematical, turns out to connect in interesting ways to broader philosophical debates about causation and explanation. The more general versions of Hamilton’s rule buy their generality at the expense of causal detail. This leads to the accusation that they explain nothing—that all the explanatory content has gone. But it depends what you mean by explanation.

As I see it, the most general version of Hamilton’s rule, though not very useful for quantitative prediction and not very causally informative by itself, serves as a kind of ‘organizing principle’ for social evolution research. It allows us to see what otherwise disparate models of the evolution of cooperation have in common: they are all models in which rb > c. And it allows us to distinguish three broad categories of causal process in social evolution: those that alter relatedness, those that alter benefit, and those that alter cost. So despite its limitations, the principle still has a pivotal role in the theory.


3:AM: How different now is evolutionary theory from the Darwinian original?

JB: In some ways, not that different. Darwinian ideas are still at the heart of evolutionary theory. But there are subtle differences. Tim Lewens’s paper ‘Natural selection then and now’ has a nice discussion of this. One difference is that Darwin often emphasises the fierceness, the brutality of the ‘struggle for existence’. It’s a vision of nature ‘red in tooth and claw’.

This makes sense given that Darwin knew next to nothing about the mechanisms of inheritance, and was trying to construct a theory that would work more or less regardless of those mechanisms. After all, if the struggle for existence is brutal, then bearers of advantageous traits will probably mate with other bearers of advantageous traits, because everyone else will be dead. Those traits, by virtue of being present in both parents, will tend to get passed on to the next generation. We don’t have to worry about the advantageous traits getting ‘blended’ or ‘diluted’ by sexual reproduction.

Now that we know a bit of genetics, we know that fierce competition is unnecessary. All you really need is differential reproduction, and the differences can be very slight. In fact, many models of social evolution assume that this is the case, i.e. that selection is ‘weak’. So the vision of nature we get from social evolution theory is perhaps a little less blood-soaked than Darwin’s. It’s one in which weak selection can drive the evolution of spectacular cooperation.

It’s a funny thing about Darwin: people seem to want to trace all the ideas of modern evolutionary theory back to him, as if he did not have enough original ideas to his name already. For example, some people want to credit him with discovering kin selection. I think this is misguided. Sometimes we see things in Darwin that aren’t really there.

We also sometimes overlook things that are there. For example, in his post-Origin work Darwin regularly appeals to ‘use-inheritance’, the ‘Lamarckian’ idea that using an ability makes you more likely to transmit it to your offspring. This is an idea that 20th Century biology went on to reject, so we now tend to gloss over this aspect of Darwin’s thought. But for Darwin, it was important.

3:AM: What are the most pressing philosophical puzzles about the evolution of human cooperation?

JB: A big question! And the central question, really, of my new project funded by the Leverhulme Trust. Where to begin? First, we need to get a grip on the things that need a special explanation: the things that make human cooperation different from cooperation in the rest of the natural world.

One obvious difference is that humans cooperate on a large scale with genetically unrelated individuals. This suggests that kin selection (at least in its traditional, genetic form) can’t be the whole story. Another big difference is the role that non-genetic inheritance (specifically, cultural inheritance) plays in sustaining human social behaviour. Accordingly, one aspect of my current project is about looking for new ways to integrate cultural inheritance into the framework of social evolution theory, and about exploring the relationship between cultural evolution and kin selection.

A quite different way to approach the puzzles of human cooperation to focus not on the evolutionary dynamics that have built and sustained it but on the psychological capacities that underlie it. There are a lot of very interesting phenomena in that category: language, learning, intelligence, trust, skill, and more. But there are two less well-known capacities that particularly interest me, and that I think are particularly central to the story: shared intention and normative guidance.

By ‘shared intention’ I mean our ability to act jointly, to do things together. When I walk to a café with a friend, we keep pace with each other, we turn corners at the same time, and this is no coincidence: we intend it. By ‘normative guidance’ I mean our ability to act in ways that are guided by norms (e.g. ‘Stand on the right! Let passengers off the train first!’) and not simply by plans or drives. My working hypothesis (for the time being, at least) is that these two capacities co-evolved, and that together they provide much of the basic psychological scaffolding for human social life.

And then the big, overarching aim of the whole thing is to connect the two halves, the evolution and the psychology: to show how cultural evolution (or perhaps ‘gene-culture co-evolution’) could produce something as amazing as shared intention and normative guidance, and to show how this in turn would have had knock-on effects on the evolutionary process. A lot to do!

3:AM: I guess there’s an issue here that can be generalised a bit – are the answers to the philosophical questions going to be solved ultimately by empirical means, and would that make the role of philosophy in all this dubious?

JB: A good question. In a way, I’m not too concerned with the issue of where science ends and philosophy begins. It’s a blurry boundary, and most of the questions that interest me seem to lie in the borderline region.

Still, one of the things that has struck me about human social evolution is how much philosophers can contribute, how much work there is for philosophers, as philosophers, to do. If we want to understand how human cooperation evolved, we need to understand what human cooperation involves, and philosophy (in particular, philosophy of action and philosophy of psychology) is essential here.

Equally, if we want to understand how the dynamics of evolution are different for our species — a species that has altered its environment like no other, giving rise to feedback loops between genes and culture that exist nowhere else in the natural world — then we need to think hard about conceptual, foundational questions (e.g. concerning concepts such as ‘cultural fitness’, ‘cultural inheritance’ and ‘cultural relatedness’) that philosophers of science are well-placed to address.

So, as I see it, the only sensible way to address these questions is through interdisciplinary work. But I don’t think this involves relegating philosophy to a subsidiary role, or that it involves setting aside philosophy’s traditional aims. The questions at stake here concern the origin and nature of the mental, social and ethical lives of human beings. These are exemplary philosophical questions. It’s simply that, in order to answer them, we need both science and philosophy, and we need collaboration and dialogue between scientists and philosophers.


3:AM: You’ve also engaged with Nick Bostrom’s ‘simulation argument’ (which sounds kind of sci-fi) and found it wanting. So what’s the issue and why don’t you think it does show that we are simulated?

JB: It doesn’t aim to show that we’re simulated—not exactly. I’ll give a brief explanation of the argument before saying what I think is wrong with it.

It starts with the idea that, given what we know about the physical limits of computation and the amount of complexity in a conscious brain, we know that a civilization with computer technology anywhere near the physical upper limit would easily be able to simulate an enormous number of conscious brains.

From this, we infer that at least one of the following three propositions is true: either (1) civilizations, including our own, are virtually guaranteed to go extinct before achieving such technology, or (2) civilisations that do achieve such technology are almost never interested in using it to simulate brains, or (3) we’re very likely to be simulated ourselves.

The idea is that, if (1) and (2) are both false, then (3) is true. The reasoning, very roughly, is that if (1) and (2) are both false then we have every reason to think that there are far more simulated brains in the universe than flesh-and-blood ones, and so a principle of indifference suggests that we’re very probably among the simulated ones.

It does sound ‘kind of sci-fi’. It does require us to be at least open to the possibility that we are living in a virtual world. This in turn requires a certain kind of picture of what a mind is. A mind must be the sort of thing that can be simulated inside a computer. Some people find that implausible. I don’t, but in any case, my real interest was in the internal structure of the argument rather than the assumptions it makes about the mind.

3:AM: So what’s wrong with the internal structure of the argument?

JB: Here’s a simple way to start to see the problem: it requires that we know things about the physical limits of computer technology, and yet proposition (3), if taken seriously, calls that knowledge into doubt. Now, that is not a problem in itself. There’s nothing wrong in principle with an argument that starts by assuming that we know something, and then shows that knowledge to be self-undermining. The real problem is more subtle. But start with that thought.

The real problem is that, to get to (3), we need to employ a principle of indifference. As I see it, principles of indifference are rarely justified. They’re justified only when you’re in a position of nearly-but-not-quite-total ignorance: you’re trying to work out whether or not you have some property, but you have absolutely no evidence to go on other than the information that x% of observers have that property.

So, when it comes to the question of whether we’re simulated or not, the use of a principle of indifference assumes that we’re in a strange evidential predicament. It requires that we know nothing at all that could help us decide whether or not we’re simulated, except all that stuff we know about the physical limits of computation that we used in order to get the argument going. I don’t know, for example, that I have two physically real human hands, but I do know a lot about semiconductors.

My claim is that our evidential predicament can’t really be like this. If our predicament is so bleak that we have to use a principle of indifference to decide whether or not we’re simulated, then we also don’t know enough physics to make the simulation argument work. Our predicament is one of pervasive scepticism, a scepticism that infects our beliefs about physics as well as our common-sense beliefs. But if our predicament is not so bleak, then we know too much for indifference-based reasoning to make any sense, and the simulation argument still doesn’t work. So either way, it doesn’t work.

But I still think it’s excellent philosophy, by the way. This is how philosophy progresses, if ‘progress’ is the right word: through bold, provocative arguments that other people come along and try to pick holes in.

3:AM: And for those here at 3:AM wanting to get further into your philosophical world are there five books you could recommend for us?

JB: I have to include something by Darwin. The Origin would be too predictable, so I’ll go for the The Descent of Man (1871). For sheer volume of ideas, it’s extraordinary. I don’t endorse all of them, particularly not when it comes to race, gender or use-inheritance! At the same time, the ratio of prescient ideas to dated ideas is remarkably high. Hilary Putnam once quipped that “as I get smarter, Aristotle, Kant, etc. get smarter as well”, and I have a similar experience with Darwin.

If Hamilton had written a book, I’d probably have to include that too, but he didn’t — unless collected papers count. Instead, I’ll go with a couple of recent philosophical books that bear on social evolution and that have helped shape my own views: Samir Okasha’s Evolution and the Levels of Selection (2006) and Peter Godfrey-Smith’s Darwinian Populations and Natural Selection (2009). Both are great entry points to debates about the evolution of cooperation.

On questions of human evolution, I recommend Kim Sterelny’s The Evolved Apprentice (2012), an excellent example of the sort of interdisciplinary work I advocated earlier. Sterelny brings insights from philosophy, anthropology, psychology and evolutionary theory to bear on questions about human evolution that no single discipline could hope to tackle alone.

In the Descent, Darwin wrote that “of all the differences between man and the lower animals, the moral sense or conscience is by far the most important”. I think he was on to something, so I also recommend Richard Joyce’s The Evolution of Morality (2006) as a readable and illuminating way in to contemporary debates about the origin of the moral sense. Joyce’s focus is on the consequences of evolution for the status of ethics: If our moral sense is an evolved adaptation, does this undermine the idea that it tracks objective moral truths? Joyce thinks it does. I think we need a clearer picture of how the moral sense evolved, and of what moral truths are, before we can settle such questions. But I also think that Joyce might be right.

Richard Marshall is still biding his time.

Buy his book here to keep him biding!

First published in 3:AM Magazine: Saturday, May 2nd, 2015.