ABSTRACT
Pedagogy is arguably humankind’s greatest adaptation and perhaps the reason for
our success as a species (Gergely, Egyed, & Kiraly, 2007). Teachers produce data to
efficiently convey specific information to learners and learners learn with this in
mind (Shafto and Goodman, 2008; Shafto, Goodman, & Frank, 2012; Shafto,
Goodman, & Griffiths, 2014). This choice not only ensures that information
lives on after its discoverer, but also ensures that information is disseminated
quickly and effectively. Shafto and Goodman (2008) introduced a Bayesian model
of pedagogical data selection and learning, and used a simple teaching game to
demonstrate that human teachers choose data consistently with the model and that
human learners make stronger inferences from pedagogically sampled data than
from randomly sampled data (data generated according to the true distribution).
Subsequent work, using the same model, demonstrated that preschoolers learn
differently from pedagogically selected data (Bonawitz et al., 2011).