Learning to Count as Algorithmic Inference (talk recording available)
Dr. Steven T. Piantadosi
Tuesday, April 15, 2014, 01:00pm - 02:00pm
University of Rochester, Brain and Cognitive Sciences
I'll describe my computational and experimental work on learning to count (Piantadosi, Goodman, & Tenenbaum 2012). I'll present a Bayesian learning model that takes naturalistic parental number usage and infers a procedure for counting. The model formalizes an approach much like Carey (2009)'s bootstrapping model, but based on rational principles of statistical inference and structured algorithmic representations. As such, the model walks a principled middle ground between strongly nativist and strongly empiricist approaches to number, showing how number knowledge might involve genuine conceptual creation, but be grounded in more basic concepts that are acquired early or innately specified. The behavior exhibited by the model matches children's developmental trajectory and stages of knowledge, while generating predictions about the fundamental processes involved in learning number. I will also discuss several current experiments on number acquisition with the Tsimane', a farming-foraging group in Bolivia. I will show that, consistent with the predictions of a rational learning model, they progress through the same stages of number knowledge as children in industrialized countries, but take longer to do so, likely due to less parental input.
Note: If you would like to receive email announcements about the colloquium series, please contact the Business Office to have your name added to our announce lists at firstname.lastname@example.org.