Wednesday, 13 March 2013

Daniel Dennett: How Darwinian is Cultural Evolution?

The blurb reads:

Dan Dennett giving a talk at the Ecole Normale Supérieure during "Sperber Week," Paris, Dec, 2012.

Dan's next talk is in the UK, titled: Cultural Evolution: from memetic evolution to intelligent design. Video, please!

Dan Sperber responds at the end of the video. He proposes that we call adaptive peaks in the memetic fitness landscape (what he calls cultural attractors) by the name "memes". He says:

Here's a simple solution: a meme is an attractor. That's what it is it's a cultural attractor.
From my perspective, this it a daft suggestion. Memes are things that move around on fitness landscapes. Attractors are peaks on the fitness landscapes themselves. These are very different ideas - not to be confused. Memes may often be found on adaptive peaks, but they can also be found elsewhere on the fitness landscape as well. You need to have both concepts. Combining them into one concept absolutely will not do.

Peaks in fitness landscapes are a commonplace idea in population genetics. For a treatment of them which relates them to memes, see here.

One of Sperber's points is that some cultural information is transformed during copying (in a way that generates attractors that precede selection). However, he doesn't seem to think this happens during copying of DNA - while, in fact, it does. DNA has its own copying distortions, as a result of error-correction gone wrong, malfunctions of the copying machinery and other factors.

Transformations during copying are a type of mutation in genetics and memetics. Memetics, like genetics, has a modular structure that allows different theories of mutation to be used. A typical model of mutations in both domains is that mutations are random. This is, of course, incorrect, but it is a helpful simplification that makes models tractable - and avoids the charge that unrestricted mutations can predict any type of change, thus rendering resulting predictions useless. Use of such simplified models in the face of complex phenomena is absolutely routine for scientists.

Dan Sperber is quite insulting about memetics, claiming that its insights are OK for philosophers, but are inadequate for empirical scientists, who must deal with culture's complexities. Basically, like most meme critics, he doesn't really understand memetics - and so isn't in much of a position to offer intelligent criticisms. What he does say shows that he has a different understanding of the concept of "copying" from the memeticists. For me, the term "copying" has a technical meaning - linked to Shannon's concept of mutual information and the concept of causality. Perhaps the idea needs spelling out for the social scientists. Having said that, Sperber (2000) did a reasonable job of explaining the concept (calling it "replication"): He wrote:

For B to be a replication of A,
  • B must be caused by A (together with background conditions),
  • B must be similar in relevant respects to A, and
  • The process that generates B must obtain the information that makes B similar to A from A.
This doesn't say what "relevant respects" means - and causality isn't always an obvious concept - but, overall, this isn't too bad.


  1. Does Sperber have any evidence that memetics is inadequate for empirical scientists, or is he just being philosophical?

    Perhaps then, we give up in the face of complexity and be content with postmodern waffle.

  2. I think Dan Sperber was just giving his perspective - as an empirical scientist. You have to be careful when arguing with Daniel Dennett, though: he's almost always right. This is a case in point: Dennett has a sympathetic understanding of memetics while Sperber does not.

  3. Hello, what was the date of this conference please? It says 'December 2013', but that is in the future. Was it December 2012 perhaps?