At this week’s MCQLL meeting, Amanda Doucette will be presenting Causal Discovery Methods for Observational Data.

Tuesday, March 26, 15:00–16:00 (Montréal time, UTC-5)
MCQLL meetings this semester are in hybrid format. We will meet in-person in room 117 of the McGill Linguistics Department, 1085 Dr-Penfield. If you’d like to attend virtually, the Zoom link is here.

All are welcome to attend.

  • Speaker:
    Amanda Doucette
    Causal Discovery Methods for Observational Data

    Many questions asked by linguists are about causal effects, although they are not always framed that way. What causes a phoneme to change over time? What causes a language to have SOV vs SVO word order? What causes children to acquire language the way they do? Does greater complexity in one area of a language cause lower complexity in another area? Despite the causal nature of these questions, linguists rarely employ causal inference methods, and often only make weak claims about causality. In many cases, designing an experiment to identify causal effects is difficult or even impossible, and we must rely on corpus data instead. Causal discovery methods are able to identify causal structures that are consistent with observational data without any experimental intervention. In this talk, I will present an overview of these methods, using data on morphological irregularity, phonotactic complexity, word length, and frequency as an example.