Thursday, August 12, 2010

Group selection: why I now think it is important

Group selection has been out of fashion for a while, subject to withering attack by Richard Dawkins and others. In general, I found the criticisms of Dawkins convincing and so missed the point of group selection. But as Jeff Schloss pointed out in his talk at the Faraday Institute that I noted last week, modern group selection theory is different from the now discredited idea that groups themselves can be selected for (and Dawkins is correct to dismiss this idea). However, there is an alternative sort of group selection that makes a great deal of sense and is fully compatible with the neo-Darwinian emphasis on the gene a the unit of selection. Today, these ideas are most associated with David Sloan Wilson, but my treatment shamelessly rips off what Jeff had to say at the Faraday.

Instead of groups, let’s think about football teams. Consider first, a rubbish team. We’ll call them England. The manager of England pays only for each goal that a player scores. As a result, all the players are desperate to score, but never pass the ball. This means that England are not very good. Consider second, a good team which we’ll call Spain. Players in this team also get paid for scoring goals, but less per score. But, additionally, the manager of Spain pays the whole team a bonus if they win, such that he expects his wages bill to be the same as England's. As a result, Spanish players pass the ball a lot to maximise the team’s goal scoring chances. This means that Spain are much better than England and each player actually scores more goals.

Now group selection says that your reproductive chances are boosted when you are a member of a successful group just as you’ll score more often if you are a member of a successful football team. Indeed, your reproductive chances are also increased if you are a member of a successful football team. In other words, the group forms part of the environment within which the individual’s genes are selected and genes that help the group will be favoured. So in Spain genes for passing the ball are favoured over those of selfishly trying to score yourself.

But as the group favours particular genes (or distributions of genes), the nature of the group itself will change over time and evolve. Just as a football team has room for strikers and midfield play-makers, so groups can accommodate different kinds of individual. This means that in a limited sense the group can be subject to natural selection, at least relative to other groups, such that it is fair to speak of group selection.

We can fruitfully speculate that altruism within the group, even altruism that does not directly favour the genes of each individual, could evolve in these circumstances. And it seems just as likely, as David Sloan Wilson has proposed, that group selection can help to account for some of the complex of behaviours associated with religion.

Discuss this post at the Quodlibeta Forum


Donny said...

Great. Now I'll never be able to look at a football match the same way again. =)

johnf said...

This reminds me of different approaches to project management with many companies working to deliver a project within time and cost.

The classical approach is via a prime contractor who then subcontracts to smaller companies who each try to meet the terms of their subcontract. With everyone fighting for themselves, this can lead to a sub optimal result.

An alternative which was used with great success by BP ten years ago (how things change!) was to enter into an alliance with the other contractors. There would be a lead contractor (BP) and subcontractors. As well as individual subcontracts with incentives there would also be a group target cost with a share line which represented an incentive fee for doing better than the target. If the project delivered inside the target cost, the reward would then be shared between the partners in agreed proportions.

Each member of the alliance would have two parameters to aim for:
a) the individual profit margin which would be small
b) a cooperative profit margin that could be substantial.

To make this work requires a lot of trust between the partners, and personal chemistry counts for a lot.

Ignorance said...

I am not sure, but didn't Professor Dawkins support some form of reciprocal altruism while denouncing group selection? What is the difference between the reciprocal altruism he champions and this strand of group selection?

The Perplexed One said...

The difference is that reciprocal altruism (as traditionally defined) is based on the exchange of favours via contract between self-interested agents (or genes, if we're going along with Dawkinsian animism). Basically think of Rousseau's concept of the Social Contract applied to the whole of nature (in other words, animals help others because they expect others to help them in return, or alternatively because their genes decieve them into helping so that their meat puppets may be protected by other genes in the future).

The problem with such a concept is that absent the language and ethical systems that regulate such contracts in humans, it is absurdly easy to "free-ride" on these interactions and constantly take advantage of favours offered (which would mean opportunists would outselect altruists). Because of this, a few scientists are very sceptical that the idea can even work, and hence seek alternative explanations for altruism.

The advantage of modern group selection is that it avoids such difficulties. It argues that where there is intense competition between groups of social animals, even though opportunists may locally be more successful within groups, they are still less likely to reproduce than members of groups that co-operate better. So, for example, if you ruthlessly exploit and deplete your own group, you will be very successful at reproducing, but you may well end up hoist by your own petard later on when a better-disciplined group tries to take your infighting offspring down.

A better example (from David Sloan Wilson) is that all of agriculture relies on group selection. A while ago a team of farmers tried to breed groups of chickens that produced the most eggs. Logically you'd think the way to do that would be to combine chickens which produce the most eggs individually, but this failed. The chickens all hoarded as much as they could, which decreased overall egg yeild from the group. You can produce much more if you select birds that get on well with each other, decreasing stress and increasing the average number of eggs.

The advantage of this approach is that it need not rely on the advanced minds that are necessary for reciprocal altruism. The agent need not be aware of it at all - even slime moulds can evolve a crude form of altruism in this way.

Ignorance said...

Okay, thank you, that was a very good explanation. To put it rather briefly, the difference would be that reciprocal altruism would require to some degree conscious agents or is otherwise liable to leeching while group selection can make cooperation an advantage in the struggle for life without such conscious checks.

PerplexedOne said...

That is correct, at least as I understand it.