Creating a Large Language Model of a Philosopher

Eric Schwitzgebel, David Schwitzgebel, and Anna Strasser

in draft

Can large language models be trained to produce philosophical texts that are difficult to distinguish from texts produced by human philosophers? To address this question, we fine-tuned OpenAI’s GPT-3 with two sets of training data: the blog posts of Eric Schwitzgebel and the works of prominent philosopher Daniel C. Dennett. The Schwitzgebel model produced extended “blog posts” that were structurally similar to the style of the philosophical blogosphere, including an extended thought experiment with structured argumentation (though of poor philosophical quality). To explore the Dennett model, we asked the real Dennett ten philosophical questions and then posed the same questions to the language model, collecting four responses for each question without cherry-picking. We recruited 425 participants to distinguish Dennett’s answer from the four machine-generated answers. Experts on Dennett’s work (N = 25) succeeded 51% of the time, above the chance rate of 20% but short of our hypothesized rate of 80% correct. For two of the ten questions, the language model produced at least one answer that experts selected more frequently than Dennett’s own answer. Philosophy blog readers (N = 302) performed similarly to the experts, while ordinary research participants (N = 98) were near chance distinguishing GPT-3’s responses from those of an “actual human philosopher”.

By following any of the links below, you are requesting a copy for personal use only, in accord with "fair use" laws.

Click here to view the final manuscript version as a PDF file: Creating a Large Language Model of a Philosopher (pdf, November 2, 2022)

Click here to view the final manuscript version as an html file: Creating a Large Language Model of a Philosopher (html, November 2, 2022)

Or email eschwitz at domain: ucr.edu for a copy of this paper.


Return to Eric Schwitzgebel's homepage.