Cultural transmission, reinvention, and progress
I sometimes wonder why some of the cleverest things about the digital age seem to date from before the digital age really started. Few philosophical analyses beat thought experiments like Turing’s imitation game or Nozick’s experience machine. Neuromancer and Wargames seem as relevant today as they were in the early 1980s. And “memes”, of course, are more than 40 years old. Perhaps one reason is that books and movies lost touch with the digital world when the internet took off: naturally enough, clever conversations about what went on online took place online; mainstream media, starting with university presses, had a hard time following. Internet studies are a thriving field, of course, but in this reader’s arrogant opinion they tended to be too specialised or too breezily generalist. It’s been all the more enjoyable, then to see some excellent recent non-fiction tackle digital culture in a way that is both scholarly, entertaining, and ambitious: books like Gretchen McCulloch’s Because Internet or Alberto Acerbi’s Cultural Evolution in the Digital Age.
I loved this book. Alberto’s nuanced and informed take on winner-take-all dynamics, fake news, echo chambers, and other internet scares, are all badly needed in a domain that has invited so many anxious pieces of lazy punditry. But this excellent book provides much more than a lucid perspective on our hopes and fears for today’s internet. It also contains an original and coherent rethinking of important theoretical issues in cultural evolution. This discussion piece focuses on the cultural evolution aspect of the book. I happen to agree with 90% of it, so I’ll focus on the remaining 10%.
Chapter 8 of Cultural Evolution in the Digital Age (CEDA) develops a hypothesis about cultural evolution (drawing on previous work from Acerbi et al., 2017; Acerbi & Mesoudi, 2015). It focuses on cultural progress through cumulative improvements, the kind of progress that got us from measurably inferior ways to produce energy, transport goods and people, or store information, to measurably more efficient ways to do all these things, in a series of incremental steps. (The word “progress” is not a word Alberto himself uses: he prefers “cumulative improvement”.)
Alberto’s argument starts from a distinction between two broad kinds of cultural transmission, the preservative and the reconstructive (p. 159–162—a distinction first proposed by Dan Sperber—Sperber, 1999). Cultural transmission is preservative to the extent that its success depends on the copying of a model with high fidelity; it is reconstructive to the extent that the content being transmitted can be retrieved without a specific model. Some cultural traditions clearly require high-fidelity copying at every step of their transmission if they are to survive at all. Sometimes we know this because they were actually lost: the recipe for Greek fire, the stone-carving techniques of the Maya, the rituals of the Eleusian mysteries. Other traditions are so odd and peculiar that they are vastly unlikely to be reinvented if they ever become lost: the exact wording of most prayers, or the self-referential use of the word “covfefe”. On the other side of the spectrum, we find traditions that are more or less easy to reinvent independently without a model: writing, Mendel’s laws, telescopes, differential calculus, are famous examples (mine, not Alberto’s).
These extreme examples are meant to fixate your ideas, but everyone who uses this distinction acknowledges that reality is not clear-cut. Rather, it points at two ideal types of transmission, with most actual cases of transmission occurring somewhere in between. The perfect way of cooking pasta is probably not something you would invent on your own—it is indeed easier to copy it (preferably from Alberto); but neither is it hard to imagine that a sufficiently motivated cook, or a community of cooks, could reinvent it, given enough time. Few behaviours are ever imitated with 100% perfect fidelity—one must leave some things out. Pure independent reinvention does happen, but it is not what people who study cultural transmission (like Alberto or me) are mostly interested in, at least when we study culture in humans. Reconstructive transmission is not the opposite of culture, it’s just a mode of cultural transmission where individual initiatives matters a lot, and the exact copying of a model matters less. We’ll delve deeper into these grey areas later on, but a simplified distinction will suffice to get Alberto’s main point. He identifies a range of domains where preservative transmission is paramount, and reconstructive transmission can only do so much. These are cases where cultural evolution solves complex problems that have few obvious solutions. One could say (my phrasing, not his) that in these situations good ideas are rare & peculiar, as well as opaque.
Good ideas are rare and peculiar means that some problems can only be solved in a few specific ways. The margin for error in making a sea-good canoe, a decent macaron, or a Windsor bow-tie (Acerbi & Mesoudi 2015) is tiny. The smallest error can ruin the technique. The only way in goes through a narrow gate. Acerbi’s technical phrase for this: the search space is narrow-peaked. To look for the solution to a problem like canoe-making is like navigating a landscape dominated by a few hard-to-reach summits. Once you reach the summit, the only way you can go is down. The thing to do, then, is to follow others who reached it before you—and stay exactly where they arrived.
Good ideas are opaque if the mechanisms linking cause and effect are partly obscure to those who use a technique (p. 29–30). We don’t understand the working of most everyday things (like light switches), as well as we think we do (Keil, 2003). We imitate gestures whose causal effects are partly intractable. The claim that most technology is opaque is easy to misunderstand, since there is a sense in which it is trivially true (no human being fully understands the full causality behind any event), and one sense in which it is clearly false (if we did not understand anything about the link between a light switch and a lightbulb being on, we would never think of using the technique to start with). Less trivially, many cultural evolutionists have argued that some techniques at least are transmitted with surprisingly little knowledge of how they work. Who gets why a Windsor bow-tie ends up looking the way it does?
When good ideas are rare, specific, and highly opaque, preservative transmission wins over reconstructive transmission (p. 195). According to CEDA, the conditions that favour high-fidelity copying also make it quite likely for cumulative progress to occur. Why? Because of the way that cultural evolutionists detect and define cumulative cultural improvement. The one sure sign that a population is capable of cumulative improvement is the existence of a behaviour that no single individual could invent on their own without access to culturally transmitted behaviours of the same type (p. 191). This criterion is the one that, for instance, primatologists use to assess whether a trait is cultural or not. It is not (I think) a very good criterion: it is devilishly hard to apply in practice, and it has arguably created more controversies than it has settled. But it is the received view, and the received view practically excludes the possibility that cultural progress could happen to ideas that are mostly transmitted by reconstruction. Cumulative improvements are not worthy of the name if they can be reinvented.
This leads Alberto to propose a distinction between two types of cultural domains, those where transmission requires copying, and where cumulative improvements may happen, and those where transmission is reconstructive, but progress is limited (p. 191–194).
Technology falls squarely in the first category (progress is common, transmission is faithful), while folklore or language belong in the second one (uncommon progress with a lot of reconstruction). On a first reading, this claim is a bit baffling. Folk tales and linguistic constructions are, after all, easily lost, seldom reinvented, and their transmission does not seem less faithful than that of scientific theorems. But that is not Alberto’s point. He focuses on features that are susceptible of incremental progress: traits that make a language easier to learn, a folk tale more memorable or more appealing, etc. Those traits, he argues, are easily reinvented. To the extent they require cultural evolution at all, that evolution happens rapidly and easily (p. 192–193, drawing on language evolution experiment). The search landscape is smooth—rolling hills, not jagged peaks—and the smart solutions are transparent. Individual cognition reaches the top on its own, without a need for cumulative improvements.
Reconstructive transmission, thus, does not matter in the domains where progress happens. Alberto’s argument here is explicitly speculative, but it reflects a widespread view among cultural evolution researchers: models that focus on reconstructive transmission must focus on domains of culture where change does not result in cumulative improvements (Acerbi & Mesoudi, 2015; Mesoudi & Thornton, 2018). Reconstruction does not explain progress.
This claim will look much more fragile, I believe, if we accept that cultural transmission and individual problem solving can interact. Culture can augment individual cognition, not just replace it. When we acquire cultural information about a problem, we do not just become better at solving this problem: we become better at thinking about it. As Pascal Boyer often says, the contribution of culture and cognition to problem is not a zero-sum affair (Boyer, 2018). Culture and cognition enhance each other.
A first consequence is that a tiny amount of cultural input can go a long way. Small cues may suffice to trigger quasi-reinventions. Alfred Kroeber (1940) coined the term “stimulus diffusion” for cases where cultural transmission takes place with very little information being transmitted: the reinvention of writing in several illiterate populations in the XIXth and XXth century, or the reinvention of porcelain in Saxony. In such cases, reinvention supplemented what imitation did not bring—the exact way of mapping images to morphemes, the exact blend of kaolin and silica—because the problems were not completely opaque. But culturally transmitted cues gave the decisive spark that made the problem tractable. The result was not just reinvention. Cumulative improvements happened too: writing extended to new languages in elegant ways, new and better types of porcelain were designed, etc. Reinvention can thus be much easier than invention.
A second consequence: culturally transmitted cues can make a problem less opaque, and solutions easier to reach. No one would expect a lone inventor to come up with an elegant demonstration for the Pythagorean theorem, from scratch. But no one had to. Once series of numbers that verify it had been spotted (three squared + four squared = five squared, how curious!), it was quite likely that somebody (or several somebodies) would try to generalise the observation. And after someone managed to demonstrate that for the first time, others were inspired to come up with increasingly elegant solutions, from Euclid’s 4-lemmas and 14-steps demonstrations to Einstein’s one-diagram proof. In the meantime, the theorem was successfully transmitted to generations of schoolchildren, not because they memorised an opaque formula, but because the proofs made an opaque problem transparent (at least for some students). The theorem could be improved because it did not have to be copied blindly.
This leads me to propose an alternative to Alberto’s view of cultural progress. In his view (to schematise), culture leads to cumulative improvements on intractable problems whose solutions are specific and difficult to find, but not for tractable problems where individual minds can figure out some solutions, unaided. I would argue, instead, that problems are not intrinsically difficult or easy: cognition and culture make them so. The shape of the problem space (smooth or jagged) is not given once and for all; neither is our level of insight into it.
What happens when problems become tractable and easier to explore? Does cumulative improvement stop? The opposite, I think, is more likely. When good ideas are easier to reinvent, they are less likely to be lost. This in itself helps cumulative progress (an idea cannot be improved upon if it’s forgotten). More importantly, tractable problems make progress easier not just for individuals, but for populations too. The fact that Pythagoras’ theorem or differential calculus were easy to reconstruct or reinvent (in a certain cultural context) also made them easier to teach and to improve upon. The received view in cultural evolution might have things exactly backwards: reconstructive transmission could be a key to cumulative improvement.
Acerbi, A., Kendal, J., & Tehrani, J. J. (2017). Cultural complexity and demography: The case of folktales. Evolution and Human Behavior, 38(4), 474–480.
Boyer, P. (2018). Minds Make Societies: How Cognition Explains the World Humans Create. Yale University Press.
Kroeber, A. L. (1940). Stimulus diffusion. American Anthropologist, 42(1), 1–20.
Mesoudi, A., & Thornton, A. (2018). What is cumulative cultural evolution? Proceedings of the Royal Society B: Biological Sciences, 285(1880).
Sperber, D. (1999). An objection to the memetic approach to culture. In R. Aunger (Ed.), Darwinizing culture: The status of memetics as a science (pp. 163–173). Oxford University Press.