A précis of ‘Cultural Evolution in the Digital Age’

When, at the beginning of 2016, I started to think about the project that would have become the book Cultural evolution in the digital age, my goal was to apply cultural evolution theory to a topic of some real-world relevance. How the diffusion of digital online media could impact the circulation and the success of cultural traits seemed an obvious choice, given my interests. On the one hand, cultural evolution is a quantitative science and I was, and am, excited about the opportunities provided by the enormous amount of data coming from social media, online texts, web searches, and the like. On the other hand, cultural evolutionists make an explicit effort of grounding their hypotheses in cognitive science and evolutionary biology. I had the impression that at least some of the works that analysed digital online dynamics were computationally sophisticated but used somewhat rudimentary models of human cognition and social interactions. Perhaps cultural evolution could have helped. As I wrote in the book with a slogan: big data and big theory.

I did not feel as having strong theoretical stances or particularly new or controversial positions to put forward, but it seemed to me that the application to the online realm of concepts from the standard toolkit of cultural evolution – in a broad sense: social learning strategies, cognitive attractors, cultural cumulation, fidelity (or lack thereof) of transmission processes – was useful enough. Somehow, the book became more interesting to write (and, I hope, to read) during the process itself, and now I suspect that at least some of the positions were relatively new or controversial. At least, quite a few people do not agree with them.

First, following the results of the Brexit referendum and of the election of Donald Trump, and possibly culminating with the Cambridge Analytica scandal, a set of tightly related ideas started to be widespread. We live in a post-truth era dominated by alternative facts; short-sighted emotions, as opposed to rational deliberation, govern now our choices. The major culprit of this state of affair is the web, and social media in particular, where fake news are more effective than truths, data-savvy villains manipulate our thoughts, and we are trapped in echo chambers, never encountering different and contrasting opinions that could help us refine our beliefs. Sure, this is a crude sketch, but it is not too far from what many newspaper articles, as well as several academic papers and books, have presented in these years.

As I was collecting material for the book, this view appeared, at a minimum, not strongly supported by data and experiments. Fake news do spread online, but not that much, and their influence on people’s behaviour seems limited. Echo chambers do exist in social media, but it is not clear whether they are stronger there than offline. Cultural evolution in the digital age can be read as a long refutation of this set of ideas and, conversely, as an exposition of many positive aspects of digital media and online connectedness that we take for granted.

Second, and related, I ended up with the need of discussing some assumptions, often not explicit but, I believe, influential, in cultural evolution. In particular, I became more and more unsatisfied of the explicative power of notions like “social influence”, “social learning strategies” and, in fact, “social learning” itself. Take, as an example, what cultural evolutionists call “prestige bias”, a tendency to copy from prestigious individuals, individuals to whom other people show signs of deference or respect. There are good reasons to copy from those people: as long as there is some correlation between skills or knowledge and prestige, copying from prestigious individuals is better than copying at random. And we do copy from them in experiments and in real life too. Up to a point, though. In the third chapter of the book I review part of a vast literature on the influence of prestigious individuals, celebrity advertisements, social media influencers, and the results go all in the same direction: the effects are limited and, especially, dependent on many other factors. It was not difficult to find examples of experiments in which participants copied from prestigious demonstrators in some conditions but not in others, of unsuccessful celebrity advertisements, and of “influencers” whose influence was mainly explained by chance.

The same can be said for the effect of popularity (chapter four in Cultural Evolution in the Digital Age) and, more generally, for the fast-and-frugal heuristics that are collected under the “social learning strategies” umbrella. Again, it is not that we do not copy popular ideas and behaviours, or prestigious people. The point is that we cannot explain the diffusion and stabilisation of some ideas and behaviours (if this is what we want to do, as I did in the book) with these concepts. Even more, I found that in many experiments of cultural evolutionists, social learning itself was used by participants to a lesser degree than what would have been optimal for the task. There are many reasons that could explain this effect, but the finding likewise points to a broad picture in which we are not easily influenceable, swayed by the last social media frenzy or by the algorithmically generated ads on our Facebook timeline. This broad picture, in turn, is consistent with an evolutionary-grounded view of human cognition and culture: if systems of exchange of information evolved and are stable, they should be, on average, advantageous for all the individuals involved. To analyse cultural evolution in the digital age we should at least start from what in the book I call “presumption of good design”.

If it is not social influence, then, how do we explain why things become successful? I analysed the case of online misinformation (chapter five), and I proposed that one way to make sense of the (relative!) success of fake news may be found in the fact that, exactly because they do not need to be true, they are less constrained by reality. For example, fake news are disproportionally negative, often contains elements of threat or disgust, and references to intense social interactions. In sum, they can be manufactured to be cognitively appealing in a way that is not possible, or not to the same degree, for real news. To understand online misinformation, we need to focus not only on presumed faults of digital online media but, on the contrary, we should understand how they are effective in bringing to us what we find interesting, whether this is true or not.

Of course, cognitive appeal is only part of the story. It would have been unfair setting a high bar for the “social influence” explanations and not for the “cognitive” ones. I have a justification, however, for my partiality. Cognitive appeal is generally weak, often overturned by other forces, and I would not bet on the prediction of the next viral meme based on it. Still, when zooming out, we should be able to recognise its effects at a large scale. If some factors are sufficiently stable, possibly because of general evolved cognitive preferences, they should drive the effects of forces that may be stronger, but of shifting directionality, like preferentially copying from prestigious individuals, or popular ideas.

I am sure the points above are familiar to the people gravitating around this website. In fact, my approach has been mostly shaped by the work of many of these people. Thus, in the last part of this précis, I would like to briefly touch on some issues that may be, if not contentious, more open to debate. Another common assumption in cultural evolution is that high-fidelity mechanisms of transmission are of crucial importance to produce and maintain culture. In Cultural evolution in the digital age (chapter seven) I discuss how this is not necessarily true for all cultural domains. The successful reproduction of a cultural trait can be due to other reasons: information is selected, transformed, selectively forgotten, so that “copying” is often only a loose metaphor when we talk about human culture. That some stories orally transmitted show startling longevity is more a consequence of the fact that their content has been transformed, ending up being appealing and more memorable and transmissible, than of a general copying ability.

However, whereas some aspects of cultural stability can be explained with this sort of convergent reconstruction, in other cases new information needs to be transmitted and preserved, and high-fidelity transmission is fundamental for that. The likelihood of me “converging” on quantum mechanics – but also of figuring out a decent recipe for lasagne – all by myself is infinitely low. One of the central ideas explored in the book is that digital online transmission provides several fidelity amplifiers that both increase the probability of new information being successfully transmitted and decrease the cost of the process. Each time we share a link on twitter we are doing something that until a few decades ago would have been akin to magic. We can find on YouTube instructional videos about practically everything, and they are especially useful for tacit knowledge: how to tie a necktie, how to open oysters, how to braid your hair (or your daughter’s, in my case).  The recipe of lasagne in my favourite website not only has a video and a step-by-step guide with pictures, but also more than 500 comments where variants are proposed, difficult passages are cleared up, and alternative ingredients discussed.

Could all that have consequences for cultural evolution? Human culture is often described as cumulative, as it grows in complexity and efficiency, drawing on the innovations of previous generations. In chapter eight, I suggest that cumulation is not a necessary consequence of cultural transmission. Some domains, such as modern technology, appear as clearly cumulative, while others, perhaps art, less. A proposal, albeit speculative, is that domains where culture is mostly supported by high-fidelity transmission are more cumulative than domains where culture is mostly supported by convergent reconstruction. In this perspective, digital online media may enhance cumulation in domains where it was previously limited. If this is true, and in which domains this happened or it will happen, is left open, but I hope it makes for an interesting research question.

Finally, there are of course many topics that I only touched and to whom I would dedicate more attention today. The role of algorithms that select information for us is only quickly discussed at the end of the book. Not surprisingly, I have a relatively optimistic take: algorithmic biases do exist, but they are an invitation to do better, more than an unsurmountable problem of digital media, also because algorithms are essential to find relevant information online. Could evolutionary approaches to culture and cognition help design better algorithms? Another aspect I feel I partly overlooked, besides some mentions in the first chapter, is the role of producers of cultural traits. Theories of cultural evolution tend to focus on receivers to explain why cultural traits are successful: why receivers select some traits and not others, why receivers find some traits appealing. The internet showed us that we seem to be more than willing to give away, apparently for free, plenty of information (without this tendency, social media would simply not exist). Knowing more about why we do so, and how we choose what to transmit, represents an essential piece of the complex answer to the question of why some cultural traits become successful and others not.