Believe it or not, this post is more than three years in the making (I can prove it with a talk in Stockholm, dated december 2019, where I was starting to put together the argument), but, as academic publishing goes, the three papers that were supposed to be part of it are only out now, so here it is. The topic is hardly new, in general and for me, and it is a classic in the so-called Paris school of cultural evolution - in fact, I may have stolen the xerox analogy from Olivier Morin’s book - but here is my version. (Maybe the last one?)
There is an apparent paradox: on one side, humans are clearly cultural animals. Everything we see, do, use, results from accumulated knowledge that we could not re-invent by ourselves. Vaccines and electric cars, sure, but also kinship systems and lasagne. Nothing controversial here. On the other side, however (here the controversial part), humans do not seem to copy others that much. We rarely blindly copy others’ actions and beliefs in the details: in general, we reconstruct them accordingly to our previous beliefs and interests. Rich domain-specific shared knowledge influences what and how we learn from others. Finally, experiments show that, even when we should - and could easily - copy from others, we do it less than what would be optimal (more on this below). How can we reconcile these two observations? In general, cultural evolutionists tend to minimise the importance of the second observation: we copy enough to keep the machinery on, and that is the important part. Or, according to some readings, evolutionary psychologists tend to to minimise the first bit: many - or more that one would think - “cultural” items and behaviours are ultimately product of evolutionary hard-coded knowledge that interacts with variable environmental inputs. Another option to solve the paradox, that I, among many others, favour, is that interactions other than copying are crucial to produce and stabilise culture.
Now, quickly, the three papers, and then why I think this is important. When reviewing the literature on conformity- and prestige-based copying for my book, I was surprised to find out not only that the usage of conformity and prestige cues was seldom the main determinant of decisions to copy, which I more or less expected, but that participants were often not copying others at all, even when this was useful and for free. (Interestingly, given the possibility of publication bias, this was coming from experiments in which the researchers were expecting participants to copy and studying something else.) In the first paper of this mighty trilogy, we reviewed various experiments where was possible to estimate if social learning was underused, correctly used, or overused, by participants, and we found that the under-usage was very robust, around ten times more common than correct or over-usage. While there may be many reasons, also specific to the experimental set-ups, for why this is the case, a safe take-home message is that it is difficult, looking at these results, to claim that we are by default avid copiers. Various later experiments seem to confirm the outcome of our review.
The two other papers are modelling works, and they both explore how mechanisms different from copying can produce culture. I already talked about the first one here, while the second one has only been published recently, after a fairly troubled existence as a preprint. The model here focuses on claims about culture in non-human primates, but the argument is more general. We show that population-level patterns that we would consider “cultural”, such as certain distributions of behaviours in different populations, can in fact be produced by non-copying mechanisms, especially the “triggering” of behaviours already-present in the individual repertoire, following the observations of that behaviour in conspecifics. Nothing controversial, really: I suspect the difficulties in publishing this model were mainly due to two factors, one possibly interesting and the other less. The possibly interesting one is that the “triggering” mechanism in the model was considered by some reviewers and readers as indistinguishable from copying. Our agents had a set of latent behaviours, and they were (roughly) expressing them with a probability equal to their frequency in the population. But, that this seems like copying was exactly our point: without “opening the box”, we can not infer from population patterns the individual mechanisms. The second, less interesting, is that we entered in a big primatology feud (on which, luckily, I do not have anything to do), and we tried to build a detailed model to interest primatologists (while an abstract model would have been cleaner in my opinion).
Finally, why do I think all this is important? From the cultural evolution perspective, I believe there is much to gain by considering other mechanisms beside copying. Do not get me wrong, I do think copying is important, very convenient to model, and useful in many domains or as an approximation, but maybe expanding the tool kit could be an important avenue for cultural evolution research, and help to make better sense of what we observe. Also, it may be, you know, correct. Second, related to my other research interest, i.e., the spread of information of in digital and online media, this view seem to me an antidote to the often-assumed idea that we are often victims of propaganda, fake news, and so on. Given how much time (and money) is spent to fight fake news, this may be a quite important issue, as the problem seems often to be not that we are too gullible, but that we are too stubborn.
Anyway, happy 2023 everyone!
