Sunday 24 May 2020

Synopsis of my replicator critique

For most traditional meme enthusiasts, memes are "the new replicators" - or are practically defined in terms of the Dawkins replicator-vehicle framework. I'm reasonably convinced by the critics that this is an unsuitable foundation for a general purpose evolutionary framework.

I've gone into this in a video/essay Against Replicator Terminology and in my 2011 memetics book, but here's a handy synposis:

Part of the problem is terminology. Dawkins (1976) defined "replicator" as follows:

"A replicator is anything in the universe of which copies are made."

One issue (1) is the "replicator"-"replicatee" distinction. You would think the appropriate term would be "replicatee" - and "replicator" would be the thing doing the copying.

Another issue (2) is copying fidelity. The "replicate" term suggests high fidelity copying. The corresponding term without that implication is "reproduce". A replica is normally a type of high fidelity reproduction. The Dawkins definition steamrollers over this distinction.

These are terminology criticisms - but my last critique is scientific - rather than just being about what words refer to which meanings. The "replicator" terminology suggests a modeling framework in which two inheritied entities are either idenitical or different. It represents a kind of binary view of inherited information. That works quite well for advanced systems of herediy because those have evolved to be digital. Nucleic acid and language are both basically digital systems which involve and use redundancy and error correction to avoid contamination by noise. However, not all systems involving inherited information are like this. In particular some of the systems I am interested in are those which exhibit Darwinian family trees in a more-or-less analog media. For example, electrical discharges, propagating cracks and fractal drainage patterns. These exhibit copying (of position and other attribuutes) with variation filtered by the environment, and a good number of Darwinian models apply to these systems. However the replicator-vehicle framework seems pretty irrelevant to their study. This consideration has the effect of displacing replicators from the foundation of Darwinism. They are still applicable to more advanced systems with error detection and correction, (but see points 1 and 2). As students of cultural evolution sometimes point out, not all cultural inheritance involves error-corrected systems such as language.

My personal preference is to just grant this point to critics of memetics and move on. The "m" word is still pretty useful, replicator terminology aside.