Brown Bag Talk: Why Computational Learning Theory Matters for Language Learning

Title: “Why Computational Learning Theory Matters for Language Learning”
Speaker: Adam Jardine, Associate Professor, Rutgers University

Abstract: In the age of ChatGPT and other large language models, it is tempting to conclude that the success of general statistical techniques obviates the need to posit domain-specific learning mechanisms in humans, such as that provided by Universal Grammar. On the contrary, when we formalize learning problems and study the conditions under which they are solvable (or not)–the subject of computational learning theory–we find that *any* successful learner must 1) restrict the space of hypotheses it is willing to consider; 2) make assumptions about how data is being presented to it; or 3) both of these.

In this talk, I review the major results that demonstrate this, starting with the seminal work of Gold (1967). I argue that not only does this perspective provide ample motivation for a Universal Grammar, but that by using these results to study the biases of human language learners, we obtain informative hypotheses for language acquisition. To this end, I also briefly outline the research of myself and my students along these lines.

This event is in person (Schmitt 104) with an virtual option on Zoom. Zoom link: https://montclair.zoom.us/j/89282900073

Snacks and drinks provided. All are welcome!

Date & Time: Thursday, September 28, 2023, 4 – 5 p.m.
Location: Conrad J. Schmitt Hall 104 – CHSS Multipurpose Room