Children as (selective) learners

Prior developmental research has offered initial empirical support for these formal models using children's exploration as an index of their inferences as learners. When an adult pedagogically demonstrates one causal function of a novel gadget (e.g., pressing a lever plays music), the model expects a knowledgeable, helpful “teacher” to provide an exhaustive demonstration of its functions; consistent with this expectation, children infer that the gadget has no other functions and modulate their explorations of the gadget accordingly (Bonawitz et al. Reference Bonawitz, Shafto, Gweon, Goodman, Spelke and Schulz2011). These results were replicated in Yucatec Mayan culture where pedagogical instruction is rare (Shneidman et al. Reference Shneidman, Gweon, Schulz and Woodward2016), further supporting the idea that these inferences are rooted in basic social-cognitive capacities rather than culturally specific teaching practices. Learners’ expectations about helpful teachers also allow children to identify unhelpful sources of information. Beyond using accuracy as a cue, preschool-aged children understand that the same accurate information can be under- or over-informative depending on the learner's prior knowledge, and evaluate others based on what they expect of a helpful, knowledgeable informant (Gweon & Asaba Reference Gweon and Asaba2017; Gweon et al. Reference Gweon, Pelton, Konopka and Schulz2014; Reference Gweon, Shafto and Schulz2018). Although selective social learning is often characterized as detecting cues that indicate when or whom to copy (e.g., a learner's own uncertainty, particular traits of conspecifics such as age or prestige; Kendal et al. Reference Kendal, Boogert, Rendell, Laland, Webster and Jones2018), these mutual expectations and mental-state inferences allow young learners to flexibly shield themselves from various forms of misinformation.

Children as teachers

When one learns primarily through copying and imitation, the decision to copy typically falls on the learner. However, when social learning occurs via communication, the real heavy lifting comes from teachers who can choose what knowledge should be passed onto the next generation. A recent study suggests that by 5–7 years of age, children make rational decisions about what is best to teach by considering what is rewarding to learn and what is more costly for learners to discover on their own (Bridgers et al., Reference Bridgers, Jara-Ettinger and Gweonin press). Although teaching benefits learners by reducing the cost of exploration and trial-and-error, teachers must be selective because it's impossible to teach everything. By prioritizing high-utility knowledge, teachers ensure that learners acquire the most critical, valuable information without the high costs of learning. Over generations, this process curates a body of cultural knowledge that is considered worthy of preserving and teaching.

Toward a more complete picture of social learning

Ironically, the cognitive capacities that support these epistemic practices are also central to Veissière et al.'s proposal: Mutual expectations and mental-state reasoning supported by Bayesian inferences and representations of expected utility (Goodman & Frank Reference Goodman and Frank2016; Jara-Ettinger et al. Reference Jara-Ettinger, Gweon, Schulz and Tenenbaum2016; Shafto et al. Reference Shafto, Goodman and Griffiths2014). In contrast to their assumptions, communication-based social learning emerges early in ontogeny (Csibra & Gergely Reference Csibra and Gergely2009), is widespread across cultures (Hewlett et al. Reference Hewlett, Fouts, Boyette and Hewlett2011), and does not always involve explicit, verbal transmission of knowledge (Gweon et al. Reference Gweon, Tenenbaum and Schulz2010). In fact, these assumptions reflect a broader issue in the field: A disproportionate emphasis on copying as the primary means of social learning (Boyd et al. Reference Boyd, Richerson and Henrich2011; Rendell et al. Reference Rendell, Boyd, Cownden, Enquist, Eriksson, Feldman, Fogarty, Ghirlanda, Lillicrap and Laland2010).

Human cultural knowledge is far more than a random collection of information. Regardless of its source – exploration, copying, communication – the evidence we observe is incorporated into a system of abstract, structured knowledge (i.e., intuitive theories) that allow us to explain the past, predict the future, and plan our own actions. Critically, the past includes our mistakes or what we didn't know, the future includes the benefit of “knowing” to our offspring, and the actions include actively sharing our knowledge with those who will benefit from it. To understand how knowledge grows over one's lifetime and over generations, we must ask how human social learning goes beyond copying, and how smart social learners become smarter teachers who willingly take on the costs of “cognitive outsourcing.”