A philosophy journal has published a largely AI-written article

Philosophy & Public Affairs, one of the most respected journals in political philosophy, has published an article most of which was written by a language model — and it did so deliberately.

Who did what?#

According to a guest post on Daily Nous, the draft was written by Anthropic's language model Claude. The thesis came from Simon Goldstein, associate professor of philosophy at the University of Hong Kong, who ran the process and edited the text.

There is no concealed use here. On the contrary, the journal's decision has the character of an experiment: how is the field going to handle AI-assisted production?

What it opens up#

Authorship in philosophy carries a different weight from most sciences. An article usually bears one name, and that name is taken to stand for that person's reasoning. When most of a text is produced by a model, several questions follow:

• Whose idea is it? If the thesis came from a human, does a machine-written text change the authorship?
• What is peer review for? Referees assess the argument, not its origin. Is that separation sustainable?
• What standard of transparency applies? At what level of contribution must disclosure be made?

A second debate#

A second argument was running through the philosophy community at the same time: the position of philosophers employed by AI companies. One discussion on Daily Nous criticised the arrangement as the arsonists volunteering for the fire brigade.

The two ends of the debate meet here. As philosophy moves into artificial intelligence, artificial intelligence is moving into the way philosophy is produced.