POLICY: AI in the review process

[eliotwrobson]

See discussion on JOSS: openjournals/joss#1483

We recently added an AI policy for submissions #344, but we didn't include any discussion about use during reviews. We should have some discussion about this and add to our existing policies.

cc @lwasser

[sneakers-the-rat]

There are at least two things here: using LLM to generate review comments (by the author and reviewers) and using an LLM when performing the review work (chatGPT analyze this software to see if its good or nah).

On the first, it can be particularly detrimental because even if the poster is just using it to "polish up their prose," it makes it so the other participants are unsure if the poster is even in the conversation or whether they are just feeding the other comments into the chatbot without reading. We will need to explicitly address the stigma re: ESL speakers feeling the need to present with polished english, which is unfortunately a real stigma - like "we would prefer you respond in your natural voice, and if you experience bias as a result of english not being your native language, report it to an editor immediately"