Meaning by Courtesy: LLM-Generated Texts and the Illusion of Content

Abstract

Click to increase image sizeClick to decrease image sizeThis article refers to:AUTOGEN: A Personalized Large Language Model for Academic Enhancement—Ethics and Proof of Principle DISCLOSURE STATEMENTNo potential conflict of interest was reported by the author(s).Notes1 The phrase riffs on the title of Baker (Citation1987).2 Cappelen and Dever’s (Citation2021) example of an AI-generated credit score rating is helpful here. Lucie’s mortgage application is rejected, due to a credit score of 550. When she enquires into what this score means, and what it has to do with her credit-worthiness, she can at best get an answer that involves a mathematical description of input data and a description of how the data is processed through the successive layers of a neural network. But this is not satisfying to her: “How, she wonders, can a bunch of mathematical transformations, none of which in particular can be tied to any meaningful assessment of her credit-worthiness, somehow all add up to saying something about whether she should get a loan?” (2021: 9; emphasis added). Note that this does not dispute the accuracy of the rating algorithm. It only asks what it could mean for the score to have a semantic content – to be saying something about her credit worthiness – given the manner in which it was generated.3 See Nagel (Citation2014, Chapter 6) for a discussion.Additional informationFundingThe author(s) reported there is no funding associated with the work featured in this article.

Ähnliche Arbeiten