Comparative Phylogenetics

As we’ve come to see memes as analogous to genes, a natural step is to borrow some of the tools from genetics and apply them to memetics. One way that biologists test evolutionary hypotheses is with comparative phylogenetics: essentially drawing a ‘family tree’ of where a species evolved from in order to compare their traits to other species. Comparative methods have a long history – Charles Darwin used differences and similarities between species as a major source of evidence – however it can be difficult to compare traits across species because their lineages are not independent. If a species develops a unique trait it may be due to an evolutionary branching event, or it could be due to the ancestors it descended from. Comparative Phylogenetic methods allow scientists to control for the likelihood of traits or combinations of traits evolving at random (assuming Brownian motion) while controlling for each species’ phylogeny (evolutionary tree).

These methods have started to be applied to memes as well as genes, in order to study the evolution of ideas in culture, measure what affects the rate of change, and to trace back present day ideas to their likely ancestors. In genetics these methods are used to provide evidence (or counter-evidence) of specific observations or rules, for example Bergmann’s rule: that body size increases in colder climates. In memetics we can use the same techniques to trace the evolution of words in languages, for example identifying why some countries call it “tea” and others “chai”: if it spread to your country by sea, it came from the ports of Fujian and Taiwan which use the coastal “te”, whereas those that received tea by land got it from the Mandarin population centers where it’s known as “chá”. Understanding what influences our culture in the past gives us insight into ways that we could influence our culture in the future. For example snowclone analysis – identifying multi-use language templates like _____ is the new _____, the mother of all _____, or _____ made simple – has the potential for identifying what clichés are likely to work in your adcopy.

As genetic information isn’t always fully available or complete, one application of these methods is reconstructing likely phylogenetic trees, so we can determine where a species lineage likely evolved from, and project forward as to what rate of change we should expect in the future. This is also useful in memetics, as memetic information isn’t always fully available or complete, especially for anything pre-dating the internet or printing press, where very little information could be economically recorded. In particular there’s a further complication with human creativity, because most artists, musicians, or storytellers are reticent to admit where they copied their ideas from, or may not even know themselves what inspired them. Studies of this kind have successfully reconstructed the evolution of folktales, tracing a direct line of evolution from modern tales back to, in some cases, the bronze age period. In other more modern cases, these methods have been used to show that adversarial creationist legislation proposed in several U.S. states all come from the same source, and the techniques they have used to ‘evolve’ tactics for getting this type of legislation passed. There have also been studies that model technology innovations as species with traits, and successfully reconstructed their evolutionary paths. If done in real-time we could sequence what ideas are due to break into the mainstream, and invest at the right time. We could also track what memes are trending, and counter them if nefarious (much like you would inoculate against a virus), or amplify them if beneficial to our cause.

Even if we have assets recorded for the full history of a domain, the problem remains cultural artefacts – books, art, movies, tweets, blogs – must be catalogued and decomposed into their component memes to be useful for analysis. This manual encoding is slow, manual, and expensive, as well as being at times unreliable, given memes may mean different things to different people based on their past experiences. This is important work however, because it we can catalogue part of the tree then look for important phylogenetic linkages, that can point us in the right direction for where to focus further cataloguing efforts. The main point of optimism we can have is that recent developments in machine learning and AI, in particular LLMs (Large Language Models) like OpenAI’s GPT-3, can radically bring down the cost of arbitrarily labeling and encoding source texts for us. In the image domain we have new tools like CLIP Interrogator, which can deconstruct what memes are likely being used in an image. Both text and image models are progressing at exponential pace, and we are soon to have video and audio models soon too. Bringing down the cost of meme sequencing is likely to have as explosive an impact on memetics, as the human genome project had on genetics.

About the Book 📕

← Back to Essays 📝