An experimental comparison of human social learning strategies: payoff-biased social learning is adaptive but underused

Analytical models have identified a set of social learning strategies that are predicted to be adaptive relative to individual (asocial) learning. In the present study, human participants engaged in an ecologically valid artifact-design task with the opportunity to engage in a range of social learning strategies: payoff bias, conformity, averaging and random copying. The artifact (an arrowhead) was composed of multiple continuous and discrete attributes which jointly generated a complex multimodal adaptive landscape that likely reflects actual cultural fitness environments. Participants exhibited a mix of individual learning and payoff-biased social learning, with negligible frequencies of the other social learning strategies. This preference for payoff-biased social learning was evident from the initial trials, suggesting that participants came into the study with an intrinsic preference for this strategy. There was also a small but significant increase in the frequency of payoff-biased social learning over sessions, suggesting that strategy choice may itself be subject to learning. Frequency of payoff-biased social learning predicted both absolute and relative success in the task, especially in a multimodal (rather than unimodal) fitness environment. This effect was driven by a minority of hardcore social learners who copied the best group member on more than half of trials. These hardcore social learners were also above-average individual learners, suggesting a link between individual and social learning ability. The lower-than-expected frequency of social learning may reflect the existence of information producer–scrounger dynamics in human populations.     

Success-biased social learning: Cultural and evolutionary dynamics

Success bias is a social learning strategy whereby learners tend to acquire the cultural variants of successful individuals. I develop a general model of success-biased social learning for discrete cultural traits with stochastic payoffs, and investigate its dynamics when only two variants are present. I find that success bias inherently favors rare variants, and consequently performs worse than unbiased imitation (i.e. random copying) when success payoffs are at least mildly stochastic and the optimal variant is common. Because of this weakness, success bias fails to replace unbiased imitation in an evolutionary model when selection is fairly weak or when the environment is relatively stable, and sometimes fails to invade at all. I briefly discuss the optimal strength of success bias, the complicated nature of defining success in social learning contexts, and the value of variant frequency as an important source of information to social learners. I conclude with predictions regarding the prevalence of success bias in different behavioral domains.     

The evolution of social learning rules: Payoff-biased and frequency-dependent biased transmission

Humans and other animals do not use social learning indiscriminately, rather, natural selection has favoured the evolution of social learning rules that make selective use of social learning to acquire relevant information in a changing environment. We present a gene-culture coevolutionary analysis of a small selection of such rules (unbiased social learning, payoff-biased social learning and frequency-dependent biased social learning, including conformism and anti-conformism) in a population of asocial learners where the environment is subject to a constant probability of change to a novel state. We define conditions under which each rule evolves to a genetically polymorphic equilibrium. We find that payoff-biased social learning may evolve under high levels of environmental variation if the fitness benefit associated with the acquired behaviour is either high or low but not of intermediate value. In contrast, both conformist and anti-conformist biases can become fixed when environment variation is low, whereupon the mean fitness in the population is higher than for a population of asocial learners. Our examination of the population dynamics reveals stable limit cycles under conformist and anti-conformist biases and some highly complex dynamics including chaos. Anti-conformists can out-compete conformists when conditions favour a low equilibrium frequency of the learned behaviour. We conclude that evolution, punctuated by the repeated successful invasion of different social learning rules, should continuously favour a reduction in the equilibrium frequency of asocial learning, and propose that, among competing social learning rules, the dominant rule will be the one that can persist with the lowest frequency of asocial learning.     

Evolution of social learning does not explain the origin of human cumulative culture

Because culture requires transmission of information between individuals, thinking about the origin of culture has mainly focused on the genetic evolution of abilities for social learning. Current theory considers how social learning affects the adaptiveness of a single cultural trait, yet human culture consists of the accumulation of very many traits. Here we introduce a new modeling strategy that tracks the adaptive value of many cultural traits, showing that genetic evolution favors only limited social learning owing to the accumulation of maladaptive as well as adaptive culture. We further show that culture can be adaptive, and refined social learning can evolve, if individuals can identify and discard maladaptive culture. This suggests that the evolution of such "adaptive filtering" mechanisms may have been crucial for the birth of human culture.     

Societal background influences social learning in cooperative decision making

Humans owe their ecological success to their great capacities for social learning and cooperation: learning from others helps individuals adjust to their environment and can promote cooperation in groups. Classic and recent studies indicate that the cultural organization of societies shapes the influence of social information on decision making and suggest that collectivist values (prioritizing the group relative to the individual) increase tendencies to conform to the majority. However, it is unknown whether and how societal background impacts social learning in cooperative interactions. Here we show that social learning in cooperative decision making systematically varies across two societies. We experimentally compare people's basic propensities for social learning in samples from a collectivist (China) and an individualist society (United Kingdom; total n?=?540) in a social dilemma and a coordination game. We demonstrate that Chinese participants base their cooperation decisions on information about their peers much more frequently than their British counterparts. Moreover, our results reveal remarkable societal differences in the type of peer information people consider. In contrast to the consensus view, Chinese participants tend to be substantially less majority-oriented than the British. While Chinese participants are inclined to adopt peer behavior that leads to higher payoffs, British participants tend to cooperate only if sufficiently many peers do so too. These results indicate that the basic processes underlying social transmission are not universal; rather, they vary with cultural conditions. As success-based learning is associated with selfish behavior and majority-based learning can help foster cooperation, our study suggests that in different societies social learning can play diverging roles in the emergence and maintenance of cooperation.     

Experimental models for testing hypotheses about cumulative cultural evolution

The rapid appearance (over evolutionary time) of the cognitive skills and complex inventions of modern humans has been attributed to “cumulative cultural evolution” (CCE), the accumulation of knowledge and skills over generations. To date, researchers have only been able to speculate about the reasons for the apparent absence of this phenomenon in nonhumans, and it has not been possible to test hypotheses regarding the mechanisms underlying it. Here, we show that it is possible to demonstrate CCE under laboratory conditions by simulating generational succession through the repeated removal and replacement of human participants within experimental groups. We created “microsocieties” in which participants were instructed to complete simple tasks using everyday materials. In one of our procedures, participants were instructed to build a paper aeroplane which flew as far as possible, and in the other, they were instructed to construct a tower of spaghetti which was as tall as possible. We show that, in both cases, information accumulates within the groups such that later generations produce designs which are more successful than earlier ones. These methods offer researchers a window to understanding CCE, allowing for experimental manipulation and hypothesis testing.     

Spread of costly prestige-seeking behavior by social learning

Mathematical and simulation models of cultural transmission in a population where individuals may differ in their social status are developed. High-status individuals are assumed to be more influential to others but no more fertile or viable than low-status individuals. Analysis of the models suggests that culturally transmitted values, beliefs, and preferences that cause individuals to engage in prestige-seeking behavior can sometimes invade the population, even if that behavior reduces the net reproductive success of the prestige seekers. It is argued that some of the seemingly maladaptive behaviors observed in human societies may be a result of cultural evolution based on the human capacity for social learning, rather than a product of the "time lag" before the evolutionary modification of the human predisposition in response to the recent drastic changes in the environment.     

The evolutionary basis of human social learning

Humans are characterized by an extreme dependence on culturally transmitted information. Such dependence requires the complex integration of social and asocial information to generate effective learning and decision making. Recent formal theory predicts that natural selection should favour adaptive learning strategies, but relevant empirical work is scarce and rarely examines multiple strategies or tasks. We tested nine hypotheses derived from theoretical models, running a series of experiments investigating factors affecting when and how humans use social information, and whether such behaviour is adaptive, across several computer-based tasks. The number of demonstrators, consensus among demonstrators, confidence of subjects, task difficulty, number of sessions, cost of asocial learning, subject performance and demonstrator performance all influenced subjects' use of social information, and did so adaptively. Our analysis provides strong support for the hypothesis that human social learning is regulated by adaptive learning rules.     

Beyond existence and aiming outside the laboratory: estimating frequency-dependent and pay-off-biased social learning strategies

The existence of social learning has been confirmed in diverse taxa, from apes to guppies. In order to advance our understanding of the consequences of social transmission and evolution of behaviour, however, we require statistical tools that can distinguish among diverse social learning strategies. In this paper, we advance two main ideas. First, social learning is diverse, in the sense that individuals can take advantage of different kinds of information and combine them in different ways. Examining learning strategies for different information conditions illuminates the more detailed design of social learning. We construct and analyse an evolutionary model of diverse social learning heuristics, in order to generate predictions and illustrate the impact of design differences on an organism's fitness. Second, in order to eventually escape the laboratory and apply social learning models to natural behaviour, we require statistical methods that do not depend upon tight experimental control. Therefore, we examine strategic social learning in an experimental setting in which the social information itself is endogenous to the experimental group, as it is in natural settings. We develop statistical models for distinguishing among different strategic uses of social information. The experimental data strongly suggest that most participants employ a hierarchical strategy that uses both average observed pay-offs of options as well as frequency information, the same model predicted by our evolutionary analysis to dominate a wide range of conditions.     

A popular misapplication of evolutionary modeling to the study of human cooperation

To examine the evolutionary basis of a behavior, an established approach (known as the phenotypic gambit) is to assume that the behavior is controlled by a single allele, the fitness effects of which are derived from a consideration of how the behavior interacts, via life-history, with other ecological factors. Here we contrast successful applications of this approach with several examples of an influential and superficially similar line of research on the evolutionary basis of human cooperation. A key difference is identified: in the latter line of research the focal behavior, cooperation, is abstractly defined in terms of immediate fitness costs and benefits. Selection is then assumed to act on strategies in an iterated social context for which fitness effects can be derived by aggregation of the abstractly defined immediate fitness effects over a lifetime. This approach creates a closed theoretical loop, rendering models incapable of making predictions or providing insight into the origin of human cooperation. We conclude with a discussion of how evolutionary approaches might be appropriately used in the study of human social behavior.     

Simulated Trust: A cheap social learning strategy

Animals use heuristic strategies to determine from which conspecifics to learn socially. This leads to directed social learning. Directed social learning protects them from copying non-adaptive information. So far, the strategies of animals, leading to directed social learning, are assumed to rely on (possibly indirect) inferences about the demonstrator’s success. As an alternative to this assumption, we propose a strategy that only uses self-established estimates of the pay-offs of behavior. We evaluate the strategy in a number of agent-based simulations. Critically, the strategy’s success is warranted by the inclusion of an incremental learning mechanism. Our findings point out new theoretical opportunities to regulate social learning for animals. More broadly, our simulations emphasize the need to include a realistic learning mechanism in game-theoretic studies of social learning strategies, and call for re-evaluation of previous findings.     

Evolution of social learning: a mathematical analysis

Social learning is an important ability seen in a wide range of animals including humans. It has been argued that individual learning, social learning, and innate determination of behavior are favored by natural selection when environmental changes occur at short, intermediate, and long intervals, respectively. Only recently, however, has the hypothesis been examined by means of mathematical models. In this paper, we construct a simple model in which each organism uses one of three genetically determined strategies--it is an individual learner, a social learner or an "innate"--and the three types of organisms are in direct competition with each other. A reduced model, involving only the individual learners and innates, is effectively linear, and we show that by solving the eigenvalue problem of this reduced system we arrive at a good approximation to the global dynamics of the full model. We also study the effect of stochastic environmental changes and reversible mutations among the three strategies. Our results are consistent with the predictions of previous studies. In addition, we identify a critical level of environmental constancy below which only individual and social learners are present.     

The problem of adaptive individual choice in cultural evolution

This paper tries to explain how individuals manage adaptive individual choice (i.e., the decision to acquire a fitter than average behavior or idea rapidly and tractably) in cultural evolution, despite the fact that acquiring fitness information is very difficult. I argue that the means of solving this problem suggested in the cultural evolution literature largely are various types of decision rules employing representations of fitness correlated properties or states of affairs. I argue that the problem of adaptive individual choice is best solved where some of these learning rule representations are socially transmitted and some are biologically transmitted.

Convergent transformation and selection in cultural evolution

In biology, natural selection is the main explanation of adaptations and it is an attractive idea to think that an analogous force could have the same role in cultural evolution. In support of this idea, all the main ingredients for natural selection have been documented in the cultural domain. However, the changes that occur during cultural transmission typically result in convergent transformation, non-random cultural modifications, casting some doubts on the importance of natural selection in the cultural domain. To progress on this issue more empirical research is needed. Here, using nearly half a million experimental trials performed by a group of baboons (Papio papio), we simulate cultural evolution under various conditions of natural selection and do an additional experiment to tease apart the role of convergent transformation and selection. Our results confirm that transformation strongly constrain the variation available to selection and therefore strongly limit its impact on cultural evolution. Surprisingly, in our study, transformation also enhances the effect of selection by stabilising cultural variation. We conclude that, in culture, selection can change the evolutionary trajectory substantially in some cases, but can only act on the variation provided by (typically biased) transformation.     

The sky is falling: evidence of a negativity bias in the social transmission of information

The method of serial reproduction has revealed that the social transmission of information is characterized by the gradual transformation of the original message. This transformation results from the preferential survival of certain types of information and the resolution of ambiguity. Here we present evidence of a bias favoring the social transmission of negatively-valenced information across multiple transmission episodes. Ninety-two, four-person chains transmitted a story containing unambiguously positive and unambiguously negative story events, along with ambiguous story events that could be interpreted positively or negatively. Analysis using mixed-effects modeling revealed the preferential survival of unambiguously negative events over positive events, and the increasingly negative resolution of ambiguous events across successive transmission episodes. Contrary to predictions, elevated state anxiety did not enhance the social transmission of negatively-valenced information. We also found that the survival of unambiguously negative story events was positively correlated with the negative resolution of ambiguous story events, reflecting a general negativity-bias in the social transmission of information.     

Rapid evolution of social learning

Culture is widely thought to be beneficial when social learning is less costly than individual learning and thus may explain the enormous ecological success of humans. Rogers (1988. Does biology constrain culture. Am. Anthropol.  90 : 819–831) contradicted this common view by showing that the evolution of social learning does not necessarily increase the net benefits of learned behaviours in a variable environment. Using simulation experiments, we re‐analysed extensions of Rogers’ model after relaxing the assumption that genetic evolution is much slower than cultural evolution. Our results show that this assumption is crucial for Rogers’ finding. For many parameter settings, genetic and cultural evolution occur on the same time scale, and feedback effects between genetic and cultural dynamics increase the net benefits. Thus, by avoiding the costs of individual learning, social learning can increase ecological success. Furthermore, we found that rapid evolution can limit the evolution of complex social learning strategies, which have been proposed to be widespread in animals.     

Culturally-transmitted feeding behaviour in primates: Evidence for accelerating learning rates

Cultural transmission implies the rapid spread of behavioural innovations when initially naïve individuals copy more informed ones. Mathematical models of transmission feature accelerating (and in most cases, logistic) rates of learning as animals that acquire an innovation provide ever increasing numbers of informers for potential learners. Conversely, non-accelerating rates have been proposed as a null hypothesis for apparent cases of cultural transmission that can best be explained by simpler mechanisms such as trial-and-error learning. Using the AIC technique for comparing models with different numbers of parameters, this paper examines the 21 cases in the primate literature where quantifiable data are available on learning rates for presumed culturally-transmitted feeding innovations. In each case, cumulative distributions over time of the frequency or proportion of individuals that acquire an innovation are compared with three accelerating functions (logistic, positive exponential, and hyperbolic sine) and two non-accelerating ones (linear and logarithmic). In 16 cases, the best fit is given by an accelerating function: nine of these support the logistic, four support the positive exponential and three, the reverse S-shaped hyperbolic sine. Individual cases often show small differences between alternative functions, but overall trends support the cultural assumption of accelerating learning rates.