Project Title "What do you mean by that?" On the Evolutionary Advantages of Idiolects in Language
Project Description: My thesis is premised on the observation that in a population, each individual has a unique understanding of language: their personal idiolect. I aim to answer why human language consists of a set of idiolects, instead of comprising a single uniform entity? And further, what advantage, if any, is conferred by this phenomenon? While it may seem that universal languages, ones where every speaker understands language in the same way, would be more efficient, human language operates within a framework of good-enough production and good-enough comprehension. This means that information distilled through communication is not complete, but instead a “good-enough” approximation of the intended message. I hypothesize that this phenomenon necessitates generalization during regular conversation, allowing greater ability to communicate in novel environments where generalization is necessary. To study this, I am utilizing Multi-Agent Reinforcement Learning to model language development in populations of universal and idiolect-based languages, and then evaluating these populations using tests of generalizability, such as performance in unseen environments and speed of integration for new agents.
Most Important Accomplishment: I’m extremely proud to have both ran a marathon and hiked rim to rim to rim in the Grand Canyon.
Reflections on the College Scholar Program: Whilst I haven’t been in the college scholar program for long, I am already impressed by both the breadth and creativity of the projects and students I work with. Nowhere else at Cornell have a been a part of such a determined, hard-working and interesting group of individuals, and the conversations I have in class shape my thinking in a way that it hasn’t yet been moulded.