Sunday, April 29, 2007

Filling in the reductionist blind spot

The chicken breeding experiment is a few years old, but it has not been widely discussed. Here's how David Sloan Wilson describes it. (Click here for all my posts about his Evolution for Everyone.)
Bill [Muir] wanted to increase egg production by selective breeding, and he tried to do it in two ways. The first method involved selecting the most productive hen from each of a number of cages [each of which contained 9 hens] to breed the next generation of hens. The second method involved selecting all the hens from the most productive cages to breed the next generation of hens. You might think that the difference between the two methods is slight and that the first method should work better. After all, it is individuals who lay eggs, so selecting the best individuals directly should be more efficient than selecting the best groups, which might include some individual duds. The results told a completely different story. …

[Using the first method Bill had effectively] selected the meanest hens in each cage, and after six generations had produced a nation of psychopaths. Inside the cage were only three hens, not nine, because the other six hens had been murdered. The three survivors had plucked each other during their incessant attacks and were now nearly featherless. Egg production plummeted. …

[Using the second method] the cage contained all nine hens, plump and fully feathered, and judging from their expressions they seemed to be having a good time! Egg production had increased dramatically during the course of this experiment.
In a note in the chapter from which this was taken, Wilson refers to his recent article "Human groups as adaptive units: toward a permanent consensus." Here's the introduction.
Foundational changes are taking place in our understanding of human groups. For decades, the biological and social sciences have been dominated by a form of individualism that renders groups as nothing more than collections of self-interested individuals. Now groups themselves are being interpreted as adaptive units, organisms in their own right, in which individuals play supportive roles.

Let me be the first to acknowledge that this new conception of groups is not really
new. A long view of scientific and intellectual history reveals that the last few decades have been an exception to the rule. The founding fathers of the human social sciences spoke about groups as organisms as if it were common sense (Wegner 1986). Before them, philosophers and religious believers employed the metaphor of society as organism back to the dawn of recorded history.

Far from robbing recent developments of their novelty, this pedigree only deepens the mystery. How is it possible for one conception of groups to be common sense for so long, for a radically different conception to become common sense, and then for the earlier version to experience a revival? A superficial answer is that ideas are like pendulums that swing back and forth. On the contrary, I believe that the organismic concept of groups will become permanently established, in the same sense that the theory of evolution has become permanently established, even if there will always be a frontier of controversy. In this paper I will attempt to show how the ingredients for a permanent consensus are already at hand.
My explanation is the same as my explanation of entities (see, for example, sections 4 and 7 of "Putting Complex Systems to Work"), and the same as my explanation of the way in which reductionism is mis-used. Reductionism tends to be misused when we pretend that by explaining how an entity functions we can dismiss the existence of the entity as an element of nature. That is, just because we (human beings) are bio-mechanical devices doesn't mean that human beings as entities can be done away with. For those familiar with computer science this is similar to saying that even though every level of abstraction can be explained by its implementation and every object can be explained in terms of computer instructions, that doesn't mean that the level of abstraction or the object itself can be ignored. I have referred to the tendency to dismiss the existence of an entity once one understands how it works as the reductionist blind spot.

Clearly the same is true of groups. Just because groups are made up of individuals doesn't mean that groups as entities can be ignored. They must be understood as having their own existence. A prediction one would make is that groups that persist as dynamic entities (i.e., as entities that persist by extracting energy from their environment) must be taken into account but groups that exist simply by virtue of being aggregations of components don't have to be taken into account.

Wilson is very much a roll-up-your-sleeves-and-get-to-work scientist. In that spirit, here's how I would test the theory I just hypothesized. Chickens in a cage, I would predict, develop a group dynamic. Even though they didn't choose to form a group, the fact that they were confined together led them to behave as a group. It was the group, then, that evolved. If one conducted the same experiment with animals that didn't interact, i.e., that didn't develop a group dynamic, I would predict that the experiment would turn out differently. Selecting the best animal in each cage would be at least as effective and probably more effective than selecting the best cage to generate the next generation.

In other words, breeding at the group level succeeds only when the groups that are being bred are real entities and not just arbitrary collections of individuals.

To complete the thought, one has to understand how groups work. In the case of chickens (and other animals) a group is a level of abstraction that is created by how the individuals interact. Clearly one doesn't breed groups. There are no group genes. But in breeding individuals one breeds the way the individuals interact when in a group. In the experiment above, the first approach bred individuals who formed dysfunctional groups. The second approach bred individuals who formed groups that functioned well. If one were breeding animals that operated purely as solitary individuals, no matter how one bred them, no group would form and there would be no group effect.

It's important to realize that there are no ant colony genes as such. The functioning of the colony depends on how the individual ants operate. But in fact, the way the individual ants operate creates a colony-level effect. So by evolving the ants to interact in a particular way, one is creating what might otherwise be referred to as the rules of the colony, i.e., the rules that the ants obey in their interactions, which in turn determine how the colony functions.

[A historical review of the recent debate about what is now called multi-level selection is available in this paper by Samir Okasha. He has just published Evolution and the Levels of Selection, which is apparently a fairly technical treatment of the subject.]

Colony-level and similar phenomena are sometimes referred to as emergence as if they were a somehow magical. Clearly, once one sees how the rules by which the ants operate produces the colony level effect, it is no longer magical. I discuss this in "Putting Complex Systems to Work" as well as in the earlier "Emergence Explained," which is available in draft form here.

See also "Group selection and merit pay at the CSU" and "Simulating group selection."

No comments: