About Rationally Speaking

Rationally Speaking is a blog maintained by Prof. Massimo Pigliucci, a philosopher at the City University of New York. The blog reflects the Enlightenment figure Marquis de Condorcet's idea of what a public intellectual (yes, we know, that's such a bad word) ought to be: someone who devotes himself to "the tracking down of prejudices in the hiding places where priests, the schools, the government, and all long-established institutions had gathered and protected them." You're welcome. Please notice that the contents of this blog can be reprinted under the standard Creative Commons license.

Friday, November 07, 2008

Pittsburgh, part I

First update, Friday morning:

I will keep this brief (because I want to go to the next session) and general, since I'd like people to get a flavor of what philosophers discuss when they get together, but obviously this isn't a technical blog aimed at professional philosophers.

So, Roughgarden -- a biologist -- gave a nice overview of the different kinds of models that biologists use in ecology and evolution, which set the tone for the following talks, given by philosophers. She explained how models are built, what kind of work they are supposed to do for biologists, and how mathematical and graphical models can be perfectly equivalent in the predictions they make about a given system.

Griesemer talked about what models in evolutionary biology mean and, more importantly, don't mean. He drew the standard distinction between statistical and causal models, and argued that measuring natural selection (which is done by statistical modeling of field data) is only a preliminary step, which does not amount to analyzing the causes of selection. The latter requires actual ecological information as well as manipulation of field conditions, to tease potential causes apart. While it is true that biologists of course know this, it is also fair to say that far too often they talk about studying selection as if that only entailed the preliminary statistical step (which is much easier to carry out).

Bouchard posed the question of what counts as an individual as opposed to an "ensemble" (a concept that includes what biologists call populations, but is in fact broader than that). He provided the example of a bioluminescent squid, which acquires its luminescence by incorporating millions of bacterial cells in its mantle. The association is usually beneficial to the squid (which paradoxically avoids predation that way, since at night predators swimming below confuse the squid for part of the starred background!). However, occasionally the bacteria take over and make the squid sick. This system is evolving either toward better integration of the two components (bacteria will become permanently associated with the squid) or toward parasitism (more sick squids). Either way, one can reasonably ask how many individuals make up the system at any given time during evolution: one (the squid, bacteria are part of the environment); a million (the bacteria, the squid is part of the environment); a million and one (bacteria + squid); or something else? Depending on the answer, different mathematical models of the evolution of the system can be set up.

Finally, Millstein took on that old biological and philosophical chestnut of what, exactly, is genetic drift. I have to premise that Roberta and I disagree on this issue, so you need to take my comments here as certainly influenced by my own views. She directed her criticisms toward a view that claims that drift is a purely mathematical construct, with no physical counterpart. Drift, according to this view, is statistical sampling error, period. I do not actually know of anyone who holds strictly to that view: if drift had no physical consequence than biologists simply would not be interested (and, frankly, neither would philosophers of science). Rather, what some people (including me) claim, is that there is ambiguity in the way biologists think of drift. On the one hand, drift is the stochastic fluctuation in gene frequencies that one expects in finite populations simply because of random variation. Think of flipping a fair coin: if you do it a small number of times you might get a deviation from the expected 50-50 ratio, not because the coin is loaded on one side, but by chance. (Roberta does not accept the exact equivalency between drift and the coin example, which many biologists use; in a discussion after the talk I think we agreed that the analogy is good enough to hold, despite some differences in the statistical properties of coin drift vs. genetic drift.) On the other hand, biologists also think of drift as "indiscriminate sampling" of genetic variants within a population: some individuals will survive and reproduce (and their genes will pass on) not because they are more or less fit than others, but just by luck. There is a huge technical literature that is pertinent here, and I don't have the time to get into further details, but suffice to say that I think Roberta and I agreed that at the very least biologists should be careful to state what they mean by "drift" in particular contexts, and whether they are talking about physical events at the individual level or statistical fluctuations at the population level. This is important because drift plays a major role in several aspects of population genetic theory in evolution.


  1. this all sounds so interesting. I think there should be more of a marriage between philosophy and science. I have a physics major friend with whom I discuss philosophy and I find that he is very good, to use buisness terms, on a micro level while I'm better on a macro level.

  2. From your comments on Griesemer, it sounds like there was little consideration of the causal modelling work done in statistics

  3. ... Oops, I must remember to finish the message before submitting it....

    The implication of the statistical work is that it is possible to infer causality without having to resort to manipulative experiments. the argument is that, given correct assumptions about what may be causal, it is possible to sort out the causal connections. of course, knowing the causal processes involves knowing the ecology.

  4. [this is on behalf of Roberta Millstein, one of the participants to the meeting quoted in my post]

    Sorry, I know this reply is really delayed -- things were crazy for me after the PSA.

    You wrote:

    "She directed her criticisms toward a view that claims that drift is a purely mathematical construct, with no physical counterpart. Drift, according to this view, is statistical sampling error, period. I do not actually know of anyone who holds strictly to that view."

    I think you've forgotten what you wrote in Making Sense of Evolution, p. 29:

    'It is worth stressing the conclusion that drift is not a process in any meaningful sense... insofar as drift is simply a name that we give to certain outcomes that are at a particular place in the statistical distribution of likely outcomes, one cannot meaningfully ask the question, "Is this outcome the result of selection or drift?"'

    It seems to me that you very clearly did endorse the view that I called DOA (drift as outcome alone). Of course, perhaps you've changed your mind since you and Jonathan wrote the book -- certainly that is your prerogative, and I would be more than happy to hear that you have!



  5. Well, except that in the book Jonathan and I distinguish between individual-level and population- (or ensamble) level phenomena. We most certainly do *not* deny that random physical events at the individual level result in statistically observable changes at the population level. All we denied was that it makes sense to construe population-level drift as a "force" in evolution. I refer the interested readers to the full chapter of Making Sense.


Note: Only a member of this blog may post a comment.