Via a friend in Vanderbilt (Hi Simon!) I got access to a recent article in New Scientist about multi-level evolution and group selection.
It is something of The rise, fall and rise of group selection. It basically goes through the many reasons by which group selection was dismissed in the 60s and how, nowadays, people start to accept that it might be more relevant than previously thought. It is interesting that the article draws from so many theoretical models (many of them based on game theory) to show how they anticipated the experimental results that were found later. According to some (quite popular) interpretation of Dawkins selfish gene, only the most strictly selfish behaviour has a chance of spreading through a population. The experiment described in the article in which prudentvirus strains could outcompete the nastier ones proves that this does not necessarily have to be the case. In general it is clear the pro ‘multiple levels of selection’ bias in the article.
My take is that Dawkins is right that anything that is detrimental to the goal of a gene to perpetuate will have slim chances of surviving in an evolutionary process. Still cooperation is known to exist and species capable of cooperation are, in many cases, more likely to thrive than those that are not. The twist lays on the fact that cooperation does not need to be detrimental for a gene (although not necessarily for the individual that carries it). First, species in which selfishness precludes collaboration and promotes explotation, selfish genes might die due to their own success. There is one typical game in game theory called the tragedy of the commons which explains how a common good can be overexploited by uncollaborative individuals with the result of everyone in the population paying a steep price for it (potentially extintion). It is also known that collaborative behaviour can emerge in a population if it benefits other individuals with related genes (kin selection), the participating individuals (the Hawk-Dove game could be used to explain that one) or if collaborative individuals manage to establish a punishment mechanism (there was a talk in Dresden by Karl Sigmund just about a month ago in which he explained this).