Has Chicago Manual of Style incorporated AI-search capabilities?

October 1, 2024 4:48 AM

Forgive my layman's possible misuse of terms. I believe some large information repositories have been indexed or otherwise cataloged and are able to take advantage of recent AI-assisted search powers. Has Chicago Manual of Style (CMOS) or other writing style guidelines [AP, Apple Style Guide, etc] taken advantage of these improved natural language search capabilities like those in Claude.ai and ChatGPT?
posted by lometogo to Writing & Language (6 answers total)

What is the question?
posted by twelve cent archie at 5:39 AM on October 1 [2 favorites]

I'm also not sure what the question is, but part of the answer may be that Large Language Models don't do search at all.
posted by heatherlogan at 5:42 AM on October 1 [5 favorites]

I can never remember off the top of my head which of the style manuals are available in electronic format, but CMOS, at least, is. The online version is well-made, and it is searchable. It does not suffer from any search deficiencies, as far as I know, such that feeding CMOS (or other guides) into an LLM would provide additional search advantage. I could see someone feeding guides into an LLM to try to provide plain language answers and circumvent some jargon, but I would expect this to lead to problems, insofar as publishers will often use... publisher jargon, not plain language.

Speaking as a librarian who regularly answers questions related to citation formatting and related issues in Chicago, MLA, and APA, I would be wary of LLM-generated answers in this area. The most common questions can be answered quickly via the Purdue OWL and other style "helper" sites. (I would say "by Googling," but Google's AI summaries are hit and miss.) My experience is that the less common questions tend to come down to things like "how do I handle X issue according to a given publication's house style," which LLM cannot help with, or "how do I cite this weird, possible very new, type of source." I could see an LLM spitting out an answer to the second kind of question that might be satisfactory, but that would present the already-classic problem of an AI presenting a potentially problematic answer in authoritative language.

It might be helpful if you explained the problems you have with the situation as it stands right now that you would expect "improved natural language search capabilities like those in Claude.ai and ChatGPT" to address.
posted by cupcakeninja at 5:58 AM on October 1 [12 favorites]

If what you mean is "can an LLM generate Chicago manual of style citations for me" the answer is...probably, sort of, but ask very carefully what you want. LLMs like ChatGPT are very good at generating formatted text (ask it for a quatrain on a particular subject, for instance). So asking an LLM to produce formatted citations of a known work is probably going to give you something that looks like Chicago Manual citations, but you would need to carefully check the actual information, as LLMs also have no source of "truth" of statements and might invent authors or other things.
posted by griffey at 6:32 AM on October 1 [3 favorites]

If what you mean is "can an LLM generate Chicago manual of style citations for me" the answer is...

...use a different tool. There are some really, really good automated citation generators out there that can produce a mostly-correctly-formatted CMOS citation for just about anything you can throw at them. My professional favorite (academic librarian here) is zbib.org which is built by the same fine open-source folks as Zotero and uses the Zotero citation engine. No muss, no fuss, no genAI, no privacy folderol, no environmental devastation, just a mostly-correct citation.

There's no need to get involved with genAI if all you need is a decent bibliography entry.
posted by spamloaf at 7:14 AM on October 1 [15 favorites]

Are you asking if any of the big style guidelines have been updated to include instructions on how to cite information obtained from an LLM AI?

I don't know what that would accomplish. You could include the prompt, but from a practical perspective unless the LLM is already providing you with citations for it's response--in which case you should just go with the original sources and cite those--I don't see how you could even cite the AI as a source in an of itself. It would be like citing Google.
posted by RonButNotStupid at 8:16 AM on October 1 [1 favorite]