Atheer Al-Khalfa

I set up a puzzling contrast using two well-known cases: Evans’ (1973) case of Madagascar and Kripke’s (1977) Leaf Raking case. Despite their apparent similarity, only the former has traditionally been thought to involve reference change. However, it’s not clear why this is so. I survey a number of influential and contemporary accounts that fail to explain the contrast in a satisfactory way. The challenge posed by cases of reference change therefore endures. Interestingly, the challenge endures for proponents of Kripke’s well-known causal picture of reference fixing, as well as for proponents of Evans’ accounts which were attempts to improve upon that picture.

Large language models (LLMs) produce certain outputs. Why do these outputs mean what they do? One might pursue a knowledge-first explanation according to which the content of those outputs is whatever maximizes knowledge of the human reading those outputs (Cappelen and Dever 2021). This paper identifies some serious challenges for that approach based on a) the tendency of LLMs to hallucinate and b) the use of certain decoding strategies such as nucleus or top-p sampling. I argue that these features of LLMs undermine the knowledge of those reading its outputs.

The essays that comprise this dissertation bring to light several foundational challenges confronting a number of influential meta-semantic theories—those that aim to explain how the contents of utterances and/or beliefs are determined. In addressing these challenges, I argue that two strategies offer a promising path forward: (1) examining the beliefs formed upon uttering or hearing a variety of complex sentences, and (2) identifying the contents that maximize knowledge across nearby possibilities. Together, these approaches help illuminate how our mental states and utterances come to represent the world.