Pittsburgh, part III

Yesterday afternoon's session was one of the most intellectually exhilarating, but the notes that follow are not for the faint of heart, though I'll try to keep things as accessible as possible for non-philosophers. 

The point of the session was to do one of those peculiar things that happen at philosophy meetings and I wish were done at science conferences: you pick a book author and ask a couple of critics to comment on his book, then ask him to respond, finally opening up the floor to a general discussion. In this case the author was Samir Okasha (Bristol), whose "Evolution and the Levels of Selection" is one of the most stimulating books on evolutionary theory I have read in many years (see my review of it: Pigliucci, M. (in press). Okasha's Evolution and the Levels of Selection: toward a broader conception of evolutionary biology. Biology & Philosophy). The commentators at the meetings were two of the most prominent philosophers of science, Elliott Sober (Wisconsin) and Ken Waters (Minnesota). I also highly recommend Sober's newest book, "Evidence and Evolution" (Pigliucci, M. (in press). Weighing the evidence in evolutionary biology (book review, Evidence and Evolution: the Logic Behind the Science, by E. Sober). Trends in Ecology & Evolution).

Sober started things out by framing major components of Okasha's book within the ongoing discussion between realists and conventionalists in philosophy. Suppose we ask what type of selection is taking place in a given instance (genic, individual, group). For a conventionalist there is always a pragmatic answer in terms of changes in gene frequencies, regardless of the actual level of selection, because gene frequencies always change as a result of selection (if there is heritability, of course). For a realist, however, the problem is to figure out at which level(s) selection actually occurs, the issue being what is the correct account of biological facts. Sober is a realist who is interested in the question of the actual unit(s) of selection.

Okasha, says Sober, thinks there is a role for limited conventionalism when it comes to the definition of "group," obviously pertinent to the issue of multi-level selection (since "group" is one of those levels). For instance, one could define groups in terms of interacting individuals or in terms of genealogy (individuals by common descent), and one gets different answers depending of which concept of group one adopts. Sober, however, maintains that only the interactionist concept of group makes sense, a realist, not conventionalist, view.

As a second example, Sober discusses Okasha's treatment of the two major approaches to mathematically model multi-level selection, the Price equation and contextual analysis (these are well known in the bio-theoretical literature, but unfortunately I don't have time to summarize them here, I gotta go to the next session...). They are alternative ways to partition causality between, for instance, individual and group level selection. While Samir prefers the contextual approach, he concludes that in fact there is no general solution to the causal partitioning problem and that either method can be used, depending on the circumstances (a conventionalist outcome). Sober, graciously citing my suggestion in the review I wrote of Okasha's book, points out that  a third alternative might be available, a mathematical description that overcomes the limitations of Price's equation and of contextual analysis, thereby offering ammunitions to the realist perspective (I think such an alternative is in the workings, and it relies on structural equation modeling).

Ken Waters opened by saying -- correctly, I think -- that Okasha's book is important not just as a treatment of multi-level selection theory, but as a discussion of how to do theory in biology in general. For Waters there are two fundamental approaches (which mirror the realist-conventionalist divide referred to above; Ken is a conventionalist): one can adopt a fundamental framework that assumes that there is one specific way in which the world works, and the task of theory is to find the universal laws or causes that make it so. The second approach he referred to as a "toolbox" where no theory can ever be complete, only partial accounts are possible, and the theoreticians focuses on whatever process happens to be of interest at a particular time.

For Waters, Samir is committed to the fundamental view, but his own (Samir's) analysis of multi-level selection theory brings him actually to endorse the toolbox approach. (Ken called this as a "creative tension" in Okasha's book, and Samir, in his reply, pointed out that that was a very nice way to refer to what others might construe as a contradiction -- who says that philosophers have no sense of humor?)

Waters' evidence for his position is, again, Samir's treatment of Price vs. contextual analysis. Since Okasha concludes that neither approach is comprehensive and that theoreticians need to use one or the other depending on the circumstances, a pragmatic-conventionalist, toolbox-like view of theory follows.

Waters pointed out a second example, which deals with the distinction between multi-level selection theory 1 (MLS1) and 2 (MLS2). Very briefly, in MLS1 the focus is on evolutionary changes in individual characters, while in MLS2 the focus is both on those and on group-level characters (Sober's discussion earlier was framed within MLS1, though classical group selection theory is framed in MLS2). Again, Waters sees Okasha as concluding that MLS1 and MLS2 are an example of a plurality of approaches to the general problem of multi-level selection, with neither being comprehensive, thus inviting a conventionalist interpretation. Notice, however, that for Samir these are two distinct processes actually occurring in nature (a more realist position, with which I concur).

Still, the Price approach and contextual analysis describe MLS1 in different ways, reaching different conclusions, which is compatible with a toolbox concept of theory (as a personal note I must say that I do like the toolbox idea, and am not too keen on universal laws in science).

Finally, what was Okasha's response to all of this? He started out by stating that he is, in fact, a realist like Sober. On the question of what constitutes a "group," however, he clarified that for him the interactionist perspective is valid, but so is one based on reproduction (not exactly the same as genealogy, as in the discussion by Sober above). He noted that MLS1 refers to interacting groups, while MLS2 deals with reproductive groups. (Which is why I think not only MLS1 and MLS2 are in fact different natural processes, but that we should not use the word "group" for both -- it has generated much confusion!)

For Samir the source of tension between the Price equation and the contextual approach originates because we want two conditions to hold for group selection: variance in group fitness, and a group effect on individual fitness. But, again, perhaps these are best thought of as separate issues which ought to be dealt with by separate theoretical instruments.

I have certainly not been able to do justice to this discussion, folks, but I hope people got at least a sense of the intellectual excitement. If anyone is seriously interested in multi-level selection theory, as a philosopher or as a biologist, they really ought to read Samir's book, as well as the special issue of Biology and Philosophy about to come out with in-depth analyses of the book.