Are humans ‘wary learners’?

I thoroughly enjoyed reading Alberto Acerbi’s Cultural Evolution in the Digital Age (Acerbi, 2019). It provides not only a much-needed corrective to overblown claims about the power of social media in influencing our thoughts and behaviours, but also a valuable contribution to the field of cultural evolution. Here I focus on Alberto’s arguments that humans are ‘wary learners’, which might be seen as contradicting central claims made by some cultural evolution researchers.

Blind copiers or wary learners?

It is fair to say that social learning – i.e. learning from others – has been given centre stage in cultural evolution research to date, while individual cognition and asocial learning have often been downplayed. For example, Herrmann, Call, Hernandez-Lloreda, Hare, & Tomasello (2007) showed that human children outperform non-human primates in social tasks, while there were no differences between species in tasks that tap non-social cognition, such as understanding of space or quantity. This effect was largely driven by social learning, with the children spontaneously and slavishly copying the experimenters, while other great apes hardly copied at all. This led Herrmann et al. to conclude that humans possess a uniquely ‘cultural intelligence’, with social learning at its core. Studies of over-imitation similarly highlight children’s propensity to copy others, often blindly (Lyons, Young, & Keil, 2007). As Alberto writes in Chapter 2, the ‘social learning strategies tournament’ not only emphasised social learning from the outset (hence the name of the tournament), but found that winning strategies copied almost all of the time (Rendell et al., 2010). More broadly, prominent cultural evolutionists emphasise how complex human technologies and social institutions emerge from beneficial lucky accidents being preserved down the generations by high fidelity social learning, such that traditional accumulated wisdom often exceeds the ability or even the understanding of any one individual (Boyd, Richerson, & Henrich, 2011; Henrich, 2015; Richerson & Boyd, 2005).

These findings and claims sometimes give the impression (or are sometimes interpreted to mean) that humans are indiscriminate social learners blindly copying others. As Alberto notes, this is similar to the impression one gets from classic social psychology studies such as Asch’s line length experiment or Milgram’s obedience studies, where people are painted as blindly conforming or obeying others.

Alberto criticises many of these claims as being over-hyped, instead arguing that humans are ‘wary learners’. We do copy sometimes, and in ways predicted by cultural evolution models (e.g. copying prestigious people or conforming to the majority), but we do this much less often than one would expect from the above claims, and we are far from indiscriminate, blind copiers. He cites Hugo Mercier, Dan Sperber and others’ work on epistemic vigilance in support of this, who argue that people frequently refuse, quite reasonably, to copy others when it comes from untrustworthy or unreliable sources (Mercier, 2017; Sperber et al., 2010). Olivier Morin has also put forward cogent criticisms of the over-emphasis of social learning in the field of cultural evolution (Morin, 2015).

Interestingly, Alberto also cites my own work where I find that participants in lab experiments, who are faced with creating a virtual artifact via either individual trial-and-error or copying other participants, copy much less than would be expected (Mesoudi, 2011). My participants would have earned more (actual, non-virtual) money had they blindly copied the best other participants. But most did not. This has been found by several other labs as well, in different tasks to mine (Efferson, Lalive, Richerson, McElreath, & Lubell, 2008; Morgan, Rendell, Ehn, Hoppitt, & Laland, 2011; Toelch, Bruce, Newson, Richerson, & Reader, 2014).

So who’s right? Is a core claim at the heart of cultural evolution theory actually incorrect? Here are a few thoughts, inspired by Alberto’s book, on how to resolve this.

Different emphasis

Perhaps this is all a matter of emphasis, part of a long tradition of scientists emphasising certain things to sell their theories. When pushed, emphasisers of social learning would surely acknowledge that people are not blind copiers and not gullible to all and every social influence (even conmen and tricksters), and emphasisers of individual cognition would acknowledge the important role of social learning and traditions: we don’t actually reinvent the wheel in each generation. Indeed, as Alberto explains, cultural evolution models indicate that a mix of social and individual learning is adaptive relative to pure individual learning or pure social learning (Boyd & Richerson, 1995; Enquist, Eriksson, & Ghirlanda, 2007).

But perhaps there is also something to resolve here. The over-imitation studies do show that children copy blindly. The social learning tournament did show that copying almost all of the time is best. My lab experiments and those of others do show that people do not copy as much as they should, if they were maximising their monetary payoff. Models show that a mix of individual and social learning is optimal, but not how much of each is optimal, or under what conditions the mix should vary.

Cultural variation

As Alberto notes, I have found that participants from mainland China tend to copy more than Westerners in the same virtual artifact task (Mesoudi, Chang, Murray, & Lu, 2015). This suggests that the aforementioned under-use of social learning could be due to experimentalists using mostly Western participants, another example of the WEIRD bias. But as Alberto also mentions, while cross-cultural variation in a trait like social learning is important to know about, it does not really help us with the issues raised above. Westerners, while WEIRD, are still human, and we still need to explain why they apparently under-use social learning. The non-Western groups in these experiments are also far from blind copiers; in my study, the participants from the Chinese mainland copied around 30% of the time, higher than the 20% or so that the British participants showed, but still much less than individual learning.

Lifespan differences

It is notable that the comparative (Herrmann et al., 2007) and over-imitation (Lyons et al., 2007) studies that reveal slavish, blind copying are both done with young children. Recent research also suggests high (but not slavish) rates of copying in adolescents (Molleman, Kanngiesser, & van den Bos, 2019). My experiments, and those of others showing lower-than-expected rates of social learning, are mostly with college-age young adults. Perhaps people show adaptive strategies of copying extensively when young, then focus on individual learning later in life once skills are acquired in order to refine those skills.

This fits ethnographic research. Demps, Zorondo-Rodriiguez, Garcia, & Reyes-Garcia (2012) found that among the Jenu Kuruba tribe of South India, the vast majority of honey collecting knowledge is acquired via social learning by the early 20s, with subsequent occasional and infrequent social learning later on in life mostly to update that knowledge. Japanese boat-builders spend six years as apprentices slavishly copying their masters, before making minor idiosyncratic innovations in order to distinguish themselves (Buckley, 2020). Lab experiments capture only a slice of a long developmental trajectory – sometimes extensive social learning in childhood, sometimes infrequent social learning in adulthood. ‘Wary learners’ were therefore not born wary, they become more wary over the lifespan. This is predicted by some models (Lehmann, Wakano, & Aoki, 2013), but has not perhaps been fully appreciated by experimentalists or in broad portraits of the field.

Experimental design

Even within experiments, how much copying is too much, or too little? Alberto’s comments made me reconsider what my own experiments tell us.

In these studies (Mesoudi, 2011; Mesoudi et al., 2015; Mesoudi & O’Brien, 2008), participants improve a virtual arrowhead on a series of trials, or ‘hunts’. There are 30 hunts per ‘season’ of hunting. Each hunt is an opportunity to change one’s arrowhead either via social learning (copying another player) or individual learning (directly altering the dimensions of the arrowhead). Social learning is payoff-biased, which means that participants can see other players’ scores, and choose to copy the highest-scoring demonstrator. There is a complex, multi-modal fitness landscape connecting the arrowhead dimensions (length, width, thickness, shape) to payoffs, with multiple locally optimal arrowhead designs varying in their maximum payoffs.

In the aforementioned study which found under-use of social learning, participants could learn either individually or socially on every hunt (Mesoudi, 2011). The extent of social learning was therefore measured as the number of these 30 hunts on which participants copied. In this experiment participants copied on average on only about 6 hunts (20%), and analyses showed that the more participants copied, the higher their score.

In earlier studies, however, all participants were forced to learn individually for the first 25 hunts, then could choose to copy (or not) only during the final five hunts (Mesoudi & O’Brien, 2008). Here, in contrast to the aforementioned findings, the vast majority, around 90%, of participants chose to copy the highest scoring other player at least once (excluding the highest-scoring player in each group, who could not copy themselves). This was adaptive: these copiers’ scores increased significantly to match those of the high-scoring player who they copied.

Why this difference? There are several potential reasons. During the 25 individual-learning-only hunts in Mesoudi & O’Brien (2008), each participant independently explored the multi-modal fitness landscape. Some happened to find the highest peak, others found lower peaks. The latter group of low-scorers could clearly see that others had done better than them, because all scores were public. They were therefore motivated to copy. In Mesoudi (2011), however, copying from the start involves uncertainty. At the start, scores are quite similar. Maybe you are actually on the right track and just need to persist with your design, and you will eventually come out on top. Without clear evidence that others are truly doing better than you, it’s hard to know for sure.

There is also an intrinsic imbalance between social and individual learning when the former is relatively free from cost and error, as it is in these experiments. Just one hunt of social learning gives you the demonstrator’s arrowhead, while several hunts (i.e. trials) of individual learning are needed to improve your arrowhead in any one dimension. In my 2011 study, I considered under-use of social learning to be a low proportion of the 30 hunts on which participants copied. If I had coded instead the proportion of participants who copied at least once, then this would be around 90%. Perhaps the latter figure is more appropriate.

All this suggests that a headline of ‘under-use of social learning’ hides many important details, which are worth thinking about. If social learning were more difficult than clicking a button, then participants would likely spend more time socially learning, much like the six-year boat-building apprenticeships mentioned above. If there were clearer evidence to the participant that they were doing poorly relative to others from whom they can copy, as there was in my earlier experiments, then there would be more motivation to socially learn. While these details may seem like arbitrary and unrealistic details of artificial experiments, real-world learning surely also varies along these same dimensions – difficulty of social learning, transparency of others’ success etc. The wariness of learners likely depends on all these details, and more.

Different domains

Alberto and I have previously argued that a lot of apparent disagreement within the field of cultural evolution can actually be seen as a focus on different domains (Acerbi & Mesoudi, 2015). I think the same probably applies to these issues. Whether social learning is useful, and therefore how much it should be relied on, depends very much on what you are trying to learn.

If it is an opaque technology (e.g. boats: Buckley (2020); violins: Nia et al. (2015); glassware: Macfarlane & Martin (2002); wheels: Derex, Bonnefon, Boyd, & Mesoudi (2019)), or a difficult-to-acquire skill or piece of knowledge (e.g. honey collecting: Demps et al. (2012); food taboos: Henrich & Henrich (2010); computer science problems: Miu, Gulley, Laland, & Rendell (2018)), then social learning is going to be crucial. These are the cases where blind copying, lengthy apprenticeships, prestige bias, conformity, and long-standing traditions are most likely to be seen. Wariness is bad, because it blocks knowledge, knowledge that can only be acquired from others.

If it is a matter of opinion, or something about which people have strong intuitions or pre-existing beliefs, or stand to be cheated or deceived, then we are likely to be much more wary. To return to the topic of Alberto’s book, much of the information exchange on the internet probably falls into this category. Social media is full of opinions espoused by people with strong beliefs but not much evidence. The internet is full of people trying to cheat others out of their money, or spreading misinformation that serves their own or their groups’ interests. Outcomes in such cases are often uncertain compared to the efficacy of technology. A bad boat sinks, but a bad opinion can be held indefinitely.

Conclusion

Is it useful to describe humans as ‘wary learners’? It’s a valuable corrective to the idea of humans as blind copiers or over-imitators. But perhaps it also falls into the same trap by generalising over all of the factors mentioned above (age, domain, motivation, culture etc.) that shape human learning strategies. Perhaps ‘flexible learner’ would be better. That said, the internet is probably a domain where wariness is extremely useful, so Alberto seems quite right to highlight our wariness in this particular context.

References

Acerbi, A. (2019). Cultural evolution in the digital age. Oxford University Press.

Acerbi, A., & Mesoudi, A. (2015). If we are all cultural Darwinians what’s the fuss about? Clarifying recent disagreements in the field of cultural evolution. Biology & Philosophy, 30, 481–503.

Boyd, R., & Richerson, P. J. (1995). Why does culture increase human adaptability? Ethology and Sociobiology, 16(2), 125–143. http://doi.org/10.1016/0162-3095(94)00073-G

Boyd, R., Richerson, P. J., & Henrich, J. (2011). The cultural niche. Proceedings of the National Academy of Sciences, 108, 10918–10925.

Buckley, C. (2020). Japanese boatbuilders and the transmission of traditional culture. Evolutionary Anthropology: Issues, News, and Reviews, 29(2), 83–85. http://doi.org/10.1002/evan.21829

Demps, K., Zorondo-Rodriiguez, F., Garcia, C., & Reyes-Garcia, V. (2012). Social learning across the life cycle: Cultural knowledge acquisition for honey collection among the Jenu Kuruba, India. Evolution and Human Behavior, 33(5), 460–470. http://doi.org/10.1016/j.evolhumbehav.2011.12.008

Derex, M., Bonnefon, J.-F., Boyd, R., & Mesoudi, A. (2019). Causal understanding is not necessary for the improvement of culturally evolving technology. Nature Human Behaviour, 3(5), 446–452. http://doi.org/10.1038/s41562-019-0567-9

Efferson, C., Lalive, R., Richerson, P. J., McElreath, R., & Lubell, M. (2008). Conformists and mavericks: The empirics of frequency-dependent cultural transmission. Evolution and Human Behavior, 29(1), 56–64. http://doi.org/10.1016/j.evolhumbehav.2007.08.003

Enquist, M., Eriksson, K., & Ghirlanda, S. (2007). Critical social learning: A solution to Rogers’ paradox of nonadaptive culture. American Anthropologist, 109(4), 727–734.

Henrich, J. (2015). The secret of our success. Princeton University Press.

Henrich, J., & Henrich, N. (2010). The evolution of cultural adaptations: Fijian food taboos protect against dangerous marine toxins. Proceedings of the Royal Society B, 277, 3715–3724.

Herrmann, E., Call, J., Hernandez-Lloreda, M. V., Hare, B., & Tomasello, M. (2007). Humans have evolved specialized skills of social cognition: The cultural intelligence hypothesis. Science, 317(5843), 1360–1366.

Lehmann, L., Wakano, J. Y., & Aoki, K. (2013). On optimal learning schedules and the marginal value of cumulative cultural evolution. Evolution, 67(5), 1435–1445. http://doi.org/10.1111/evo.12040

Lyons, D. E., Young, A. G., & Keil, F. C. (2007). The hidden structure of overimitation. Proceedings of the National Academy of Sciences, 104(50), 19751–19756.

Macfarlane, A., & Martin, G. (2002). Glass: A world history. University of Chicago Press.

Mercier, H. (2017). How gullible are we? A review of the evidence from psychology and social science. Review of General Psychology, 21(2), 103–122. http://doi.org/10.1037/gpr0000111

Mesoudi, A. (2011). An experimental comparison of human social learning strategies: Payoff-biased social learning is adaptive but underused. Evolution and Human Behavior, 32(5), 334–342. http://doi.org/10.1016/j.evolhumbehav.2010.12.001

Mesoudi, A., Chang, L., Murray, K., & Lu, H. (2015). Higher frequency of social learning in China than in the West shows cultural variation in the dynamics of cultural evolution. Proceedings of the Royal Society B, 282, 20142209.

Mesoudi, A., & O’Brien, M. J. (2008). The cultural transmission of Great Basin projectile point technology I: An experimental simulation. American Antiquity, 73(1), 3–28. http://doi.org/10.2307/25470456

Miu, E., Gulley, N., Laland, K. N., & Rendell, L. (2018). Innovation and cumulative culture through tweaks and leaps in online programming contests. Nature Communications, 9(1), 2321. http://doi.org/10.1038/s41467-018-04494-0

Molleman, L., Kanngiesser, P., & van den Bos, W. (2019). Social information use in adolescents: The impact of adults, peers and household composition. PLOS ONE, 14(11), e0225498. http://doi.org/10.1371/journal.pone.0225498

Morgan, T. J. H., Rendell, L. E., Ehn, M., Hoppitt, W., & Laland, K. N. (2011). The evolutionary basis of human social learning. Proceedings of the Royal Society B, 279, 653–662. http://doi.org/10.1098/rspb.2011.1172

Morin, O. (2015). How traditions live and die. Oxford University Press.

Nia, H. T., Jain, A. D., Liu, Y., Alam, M.-R., Barnas, R., & Makris, N. C. (2015). The evolution of air resonance power efficiency in the violin and its ancestors. Proceedings of the Royal Society of London A, 471(2175), 20140905. http://doi.org/10.1098/rspa.2014.0905

Rendell, L., Boyd, R., Cownden, D., Enquist, M., Eriksson, K., Feldman, M. W., … Laland, K. N. (2010). Why copy others? Insights from the social learning strategies tournament. Science, 328(5975), 208–213.

Richerson, P. J., & Boyd, R. (2005). Not by genes alone. Chicago: Univ. Chicago Press.

Sperber, D., Cl’ement, F., Heintz, C., Mascaro, O., Mercier, H., Origgi, G., & Wilson, D. (2010). Epistemic vigilance. Mind & Language, 25(4), 359–393. http://doi.org/10.1111/j.1468-0017.2010.01394.x

Toelch, U., Bruce, M. J., Newson, L., Richerson, P. J., & Reader, S. M. (2014). Individual consistency and flexibility in human social information use. Proceedings of the Royal Society B, 281(1776), 20132864.

Comments Disabled

  • comment-avatar
    The ICCI Team 10 June 2020 (15:15)

    Save your comments for the discussion!
    The Book Club discussion will take place from June 29th to July 12th, and will be open to all.