Tuesday 27 May 2014

Analogies, evolutionary forces, and a shiny new journal

Analogies are all the rage in Philosophy of Science of late, doncha know. Griffiths and Stotz talk about them, Dennett talks about them (I love that Amazon has his new book filed as 'self-help' btw), and now Bristol are going to be talking about them for two days straight.

This week we can all enjoy the hotly anticipated  debut of one of Philosophy's first open-access general journal Ergo, for which we have the fantastic Franz Huber and Jonathan Weisberg of Toronto to heartily thank.
Of course, we already have the Philosopher's Imprint, which provides totally free refereed journal articles. But Ergo is the very first to try and include section editors from all areas and traditions within philosophy. They can further boast a triple-anonymous review policy and will publish data about their submissions and turn-around times.

Spearheading its launch comes a fine paper scrutinising an analogy which has been getting a bad rep in philosophy of biology lately. Velasco & Hitchcock's Evolutionary and Newtonian Forces asks how far we can go by comparing natural selection, drift and other evolutionary phenomena to physical forces such as gravity.

Against defenders of the statistical view of selection, for whom evolution is just the net aggregate of everyday birth and death events, rather than being the outcome of any overarching force, the authors claim that the analogy is in fact sound, so long as we let go of gravity being the sine qua none of dynamical forces. If, instead, we consider forces such as friction and elasticity, then we see that those features, whose absence supposedly disqualifies selection and drift from being Newtonian, are actually lacking in some of Newton's paradigm forces too. The criticisms, in other words, "rest on false presuppositions about what a theory of forces must look like." This is good news for all those, such as Elliott Sober, who have long argued that natural selection is a genuine cause of evolution, because our notion of causality finds perhaps its strongest exemplars in good old-fashioned Newtonian causes of motion.

Our chat about natural selection gets metaphorical in various ways - genes as agents, traits as designed, etc. One of the most central and, for me, vivid figures of thought casts natural selection Herself (She is always female) as an agent. I think of Her as being much like the wispy magical hand from the national lottery adverts, coming down and choosing organisms to be the lucky ones that make it. We could equally spell the idea out as some sort of giant boot that squashes the not-good-enough (Final destination style?) but I prefer the happy version. Either way, natural selection acts, She chooses, She favours, etc etc.

The notion under discussion in Velasco & Hitchcock's paper isn't this nouveau-folk-biological fantasy, however. Neither is it the sieve or funnel that sometimes makes it into evolutionary metaphor. If I try to picture what an evolutionary force looks like, it ends up being something invisible, a force field or laser beam that sort of sucks particles in a particular direction. Good job I wasn't writing this paper. Apparently gravity is the force least analagous to natural selection.Velasco & Hitchcock cut through all this nonsense and pin down a careful characterisation of a theory of evolutionary forces in terms of various kinds of law, before showing why there is no need to turn instead to a rival analogy, in which selection is recast as a mere figure on a data clerk's excel spreadsheet. For those of the statistical view, She is the life expectancy term, the estimated rate of return on an investment. She is nothing, in other words, over and above our best guess about the aggregated effects of millions and billions of run-of-the-mill causal events. Not even a creepy giant hand.

More precisely, these critics (in particular Denis Walsh, Tim Lewens, Mohan Matthen and André Ariew) hold that "The analogy between Newtonian forces and evolutionary 'forces' such as natural selection, drift, mutation and migration is misleading." Against this, Velasco & Hitchcock maintain that there is no danger of being misled, so long as we are careful to match up evolutionary forces to the appropriate Newtonian analogs. Natural Selection, for example, matches up rather nicely to friction or to the elastic force, while migration and mutation approximate gravity. The authers give a careful, detailed refutation of several supposed sources of disanalogy, including;
  • Isolability - whether or not the force can be conceived as acting independently of others.
  • Source laws - which spell out the way in which the force acts independently of any body actually being acted upon.
  • Composition of forces - whether we can compute the net effects of forces acting in concert (Sober 1984).
  • Tertium quid - whether the process in question is a distinct causal variable acting as an intermediate link in a causal chain. 
In each of these cases, they argue that those criticisms aimed at creating a distance between our picture of evolution and our picture of Newton's physical forces stem from misconceptions about Newtonian forces. For example, it's true that natural selection isn't really a separate variable that intervenes somewhere in between heritable variation in a trait and the death of some animal. But it's equally wrong to assume that gravity is some extra step that follows the appearance of two heavy bodies situated at some distance from one another but precedes the bodies' attraction of one another. Both examples fail Woodward's manipulability test of causation, in any case.

The source law requirement is satisfied by any zero force law - usually the Hardy-Weinberg equilibrium, but Brandon & McShea's ZEFL would work too. NB there are a lot of laws required, for a subject which is often claimed to involve none. And as someone who thinks mostly about case studies involving either plants or bacteria, I feel a bit twitchy about anything that puts fundamental importance on the Hardy-Weinberg equilibrium, which only applies to diploid organisms. I wonder if Lewontin's conditions couldn't act as a general source law for evolutionary forces?

One outstanding wrinkle in all this: while it's true that the argument is often framed as being about whether natural selection is comparable to a Newtonian force, and true that friction and the elastic force are definitely Newtonian forces, there might be a legitimate sense in which a defender of a statistical view could say 'but those forces are different.' As Velasco & Hitchcock acknowledge, friction and elasticity are usually thought of by physicists as emerging "from the aggregate statistical behaviour of more elementary forces in certain kinds of system." It's easiest to see this with friction - roughly friction is what we call the resistance offered to a body by the stickiness of some surface. The stickiness of the whole surface is a sum of the stickiness of the molecules from which it is composed, which ultimately boils down to weak and strong nuclear forces between subatomic particles. So in this sense we might say that friction supervenes on more basic forces, rather than having any causal oomph of its own.

But this is grist to the statistical view's mill, we might say. They argue that natural selection supervenes on more basic causal events, without adding any extra causal power of its own. So these critics might happily accept that evolutionary forces are analagous to non-fundamental Newtonian forces, whilst holding their ground on the claim that natural selection is not causal.

Velasco & Hitchcock's response to this possibility is interesting. They point out that actually "the situation is a bit tricky, since the details of how these forces arise are not completely understood. Moreover, the explanation will likely take us out fo the realm of Newtonian it may be fair to say that from the perspective of elementary Newtonian mechanics, friction and spring forces are basic."

I suppose what the authors want to do here is put pressure on the idea of some level being more basic or elemental than another. If we mean epistemically, then friction has priority over electromagnetic forces, since it is better understood. Further, we could argue for friction as emergent rather than being reducible to lower level forces, in so far as our lack of understanding about how the electromagnetic forces sum makes us unable to compute the friction they give rise to. More simply, however, there is no point complaining that selection isn't force-like on the grounds that there is something metaphysically ambiguous about its status in relation to the lower-level events with which it is intimately related, given that analogous metaphysical concerns accompany the more-and-less fundamental physical forces. Gravity is the paradigm spooky action-at-a-distance. Quantum entanglement, wtf? If there is anything magical about thinking of natural selection as an overall force producing all the multifarious births and deaths that we actually observe, then it is in very good company lumped in with physical forces.

When we compare natural selection, drift, mutation and migration to the strange, complex and only incompletely understood forces of actual physics, as opposed to the simplified forces of folk physics, the obvious disanalogies evaporate. In fact I'd guess that the straw man is actually what my A-level physics teacher, the excellent Mr Drum, used to call 'balls in a box.' The simplest kind of mechanics, the kind most people have little trouble making predictions about, concerns the behaviour of snooker balls. In Newton's mechanistic paradigm, all physical events were to be understood as arising out of the collisions of corpuscles. His laws act on these corpuscles much as snooker cues push snooker balls around on a table. Evolution can similarly be conceptualised in terms of the motion of particles, where the particles are organisms whose 'collisions' - their fitness-affecting interactions, in other words, form macro patterns. Evolutionary forces dictate the eddies and currents, some sections of the stream moving faster than others, some swirling into deadends, others branching off.

Velasco & Hitchcock say "The analogy succeeds, in part, because Newtonian forces and evolutionary forces are both heterogeneous. Gravity is very different from friction, and natural selection is very different from migration." I would add that it succeeds because actual physical forces, unlike folk physical forces, are actually pretty ambiguous and badly understood. In both cases, the notion of a force as some sort of power posessing a magnitude and a direction, like a gust of wind or the slap of a hand, is a heuristic by which we attempt to get a handle on the messy underlying reality.


  1. Thanks for the nice comments Ellen!

  2. Hi.

    Good review of the paper, but I take issue with your appeal to the unknown as a defense of V&H's causalist position. You said:

    >If there is anything magical about thinking of natural selection as an overall force producing all the multifarious births and deaths that we actually observe, then it is in very good company lumped in with physical forces.

    Explaining something unknown by appealing to something even less understood is not a good strategy (ignotum per ignotius). Let me explain why this is really problematic:

    Imagine a statisticalist pointing to their analogies and explanations of evolutionary phenomena and saying, "Evolution isn't mysterious at all, and we have a perfectly good statistical explanation right here. The only causality is in the underlying fundamental physics." The evolutionary causalist is then in the uncomfortable luddite position of insisting, without reason, that we don't understand evolution. Appealing to an analogy with physics that supports the causal position is question begging, if there is no deeper reason why this analogy holds other than it supports the claim that evolutionary phenomena are mysterious and hence causal. Therefore without some other reason to support the causal view of evolutionary phenomena, appealing to mysteriousness does not justify the causalist position.

    So I don't think your defense of their position is viable, and hence V&H's argument is on shaky ground. Though please feel free to poke holes in what I said.

  3. It's funny, but I can't find where Sober says that evolutionary theory is analogous to a theory of forces. The title of his chapter suggests that one can treat it as a theory of forces. The closest I can find is on page 14: "The theory may be thought of in analogy with Newtonian mechanics. Various forces are described, but the theory has at its conceptual center a view of what will happen to the systems it describes when no forces at all impinge." It was Chris Stephen's who said, discussing my and Andre's piece, that "the analogy between evolutionary theory and Newtonian mechanics was never intended to be perfect. However that may be, Sober always assumed that there was a clear sense of force, and that evolutionary theory is quite literally a theory of forces. This is actually what the quote from Sober I have just given seems to presuppose.

    A more substantial point. Hitchcock and Velasco say that friction never acts by itself, but always in conjunction with other forces, such as the force of gravity that pulls an object down on another, thus creating friction. "This undermines the claim of Matthen and Ariew that Newtonian forces (unlike evolutionary forces) can always be considered in isolation." This is a welcome clarification. However, we never meant to say that all Newtonian forces can act in the absence of other forces. (I doubt we even ever said that, though I'd have to check: we never discussed the point, so it's possible we were careless.) Rather our point was that in evolutionary theory, you cannot disaggregate the contributions that the various Sober-forces make to an individual evolutionary change. So you CAN say that friction was one-fifth the force of gravity, but you cannot say that in the evolution of the human opposed thumb, drift accounted for one-fifth of the final result.

    Finally, I note with some bemusement that Hitchcock and Velasco list Brownian motion as a Newtonian force. Brownian motion is a paradigm for our critique. Like drift, it is the product of uncertainty in an ensemble, not a cause of it.

  4. Noah, thanks for your comments.

    (BTW, I won’t claim to speak for Joel in anything I say here.)

    Let’s distinguish two different claims:

    1. Evolutionary forces such as selection, drift, migration, and mutation are analogous to Newtonian forces such as gravity, electrostatic attraction/repulsion, friction, and spring forces.

    2. Evolutionary forces are causes of changes in gene frequencies in a population (and they can sometimes cause stasis as well).
    In our paper, we argue for 1 (although we demur to a considerable extent on drift). We don’t argue directly for 2, although many will think that 1 lends strong support for 2.

    So suppose someone makes the following argument (I don’t attribute this to you, but maybe it is in the ballpark of what you are considering):

    i. Natural selection is more closely analogous with friction than with gravity.

    ii. Friction does not cause deceleration and heating, since all of the causation occurs at the level of fundamental forces.

    iii. Therefore, natural selection does not cause changes in gene frequencies, since all of the causation goes on at a lower level.

    Nothing we say in the paper is intended to refute this argument; in particular, we don’t challenge premise (ii). In fact, I would be happy if someone making this argument cited our paper in support of premise (i). Personally, I don’t find premise (ii) very compelling. (I’ve written about this briefly in the context of the causal exclusion argument in philosophy of mind.) And I don’t read any of WALM as being skeptical of micro-causation generally. But nothing in the present paper is aimed at refuting premise (ii).

    We are claiming that evolutionary biology is analogous to Newtonian mechanics in certain respects. Newtonian mechanics is not literally true, so in saying this, we are saying that the story about the world that is told by evolutionary biology is like the story about the world told by Newtonian mechanics (regardless of whether the latter is true). One problem with saying this is that “Newtonian mechanics” is much less clearly defined than most people realize. It is not a formal system with a fixed list of axioms like ZFC. And there are lots of branches of classical physics that loosely fit together: continuum mechanics, rigid body mechanics, Lagrangian and Hamiltonian mechanics, etc. I think that those who make an analogy between evolution and Newtonian mechanics have in mind something like: the physics you learn in a first year university physics course. This is classical mechanics with discrete bodies and discrete forces.

    If you look at the way friction is presented in such a course, it is essentially treated as an autonomous force. There may be a hand-wavy gesture at the possibility of explaining friction in terms of more fundamental forces, but this is not in any way a central theoretical commitment of the theory being presented. So the main claim we wanted to make here is that the reducibility of friction is not an essential part of the story NM tells about the world. A possible world in which friction is primitive is still a Newtonian world. Friction, whether reducible or not, is very much a force in the context of NM.

  5. Mohan, thanks for your comments.

    1. I agree with you about what Sober says. He seems to think there is a sense of ‘force’ such that we can talk about theories of forces other than Newtonian Mechanics. NM is the most familiar theory of forces but not the only one. I agree with what I think you are suggesting: there isn’t a clear sense of force outside of the context of NM. In particular, I would have a hard time making sense of someone who said: “Theory X is a theory of forces. Even though ‘forces’ in theory X are almost completely different from forces in NM, they still have property Y, which is the essential feature of forces.” I don’t know what could tell us what the essential properties of forces are if not NM. So you were certainly right, in challenging Sober, to try to identify differences between evolutionary forces and Newtonian forces. I also agree that the dynamical view has evolved on these issues, in part in response to your criticisms.

    2. In the text, we don’t claim that Brownian motion is a force. We included it in the diagram just to indicate what we thought the closest Newtonian analogs of various evolutionary forces are. We say in the text the drift is the least force-like of the alleged Newtonian forces, and that there are important issues about the role of probability in evolutionary biology for which the analogy with Newtonian forces is unenlightening. So we are not claiming that all evolutionary forces are perfectly force-like. Part of our point is that some of the most force-like are the ones receiving relatively little attention, like mutation and migration.

    3. Part of the point that we were making in the section on composition of forces is that you can’t desegregate the over-time effects of individual Newtonian forces either. E.g. if a test particle, experiences gravity due to two masses m1 and m2, and after ten seconds it has deflected 10 meters south of where it would have been in the absence of forces, you can’t decompose that 10 meters south into components due to m1 and m2. (As we note in the paper, this point is made very nicely and in more detail in a paper by Sheldon Smith.)

  6. Hi Chris,

    Thanks for your response. In the interests of not getting into an endless back and forth, I'll restrict my comment to something new that arises out of your reply.

    Here is a heuristic way of putting my point about selection and drift. Selection is a chance process; drift is the departure from expectation in that chance process. To suggest that such departures require an additional cause ("drift as process," as Roberta Millstein calls it) is to suggest contrary to assumption that selection is deterministic and departures are explained by interfering causes. (Sober does actually say something like this, but he has to add the caveat that selection is deterministic in infinite populations. True, but there's also no drift in infinite populations.)

    Your point about mutation and migration being closer to Newtonian forces is interesting. Mutation and migration can be disaggregated post facto, and this is an important point. The only problem in the general theoretical context is that these too are probabilistic. Mutation figures in the mathematical theory as a mutation rate, and I would argue that you can't disaggregate mutation rates from selection even post facto.

    A further complication—and I am not sure as yet what to make of this—is that mutation and migration rates figure in the selection process. For instance, a propensity to increase mutation rates when stressed may be adaptive; similarly for migration.

  7. Following the same point......
    Mohan you said "you cannot disaggregate the contributions that the various Sober-forces make to an individual evolutionary change."

    I wonder if Price's equation doesn't give us one way we can disaggregate component forces of the total evolutionary change of a population. Steve Frank has a nice series of papers in Journal of Evolutionary Biology explaining how we can separate the evolution that is due to natural selection from the evolution that is caused by transmission bias. Admittedly, this doesn't exactly line up with a decomposition into selection, drift, mutation and migration. Rather, the transmission bias term incorporates mutation, lower-level selection and drift. But we can separate these out in models at least, holding one or more of them constant.

  8. Here's a link.....

  9. Hi Ellen,

    I worked quite hard on the Price equation way back when André and I were working on our 2002 paper. At the time, I concluded that it was not fundamentally different with respect to this point than Fisher's theorem or general descriptions. My position has always been that you cannot separate out the contributions of these factors in individual histories. You can do so for the type of situation that is presented by a population in a certain situation. In other words, my claim is that it is one thing to say that 65% of the time, a situation such as that faced by the hominin population 500,000 ya, language would evolve (or that this is the probability of this outcome) and another thing to say that in the actual evolutionary history of human language, selection was 65% responsible. I actually don't think the second has any clear meaning. A quick look at Steve Frank's paper leads me to think that he is talking about mutation rates and selection strength. I could be wrong, of course, and please correct me if I have missed something essential, but the paper doesn't seem to be new with regard to the point in question.

  10. Hi Chris,

    I find your claim that "our concern in the present paper is with the analogy for its own sake" [from the introduction to the paper] to be problematic.

    If the concern is the analogy for its own sake, then the conclusion is an accurate sociological/ historical/ scientific description of how we talk and think about evolution currently. But this isn't a philosophically innocent endeavor. There is the divide between the statisticalists and the causalists, and I suspect each thinks about evolution in significantly different ways from the other. So I doubt there is an independent view of the analogy outside some causalist/ statisticalist bias. It is a loaded issue.

    So what I was responding to above was that I got a bit of a causalist bias since you "are saying that the story about the world that is told by evolutionary biology is like the story about the world told by Newtonian mechanics". I took this to be a biased position against the statisticalists, and was saying that the statisticalists could provide analogies that would tilt evolutionary biology their way. Everyone can insist their analogies are the right analogies, but, without further argument, this is only to reiterate one's basic position.

    Appealing to the amorphous nature of physical forces, as I took EC to be saying, won't help solve this problem.