There is also lots of national news coverage (Post, Globe, etc), though mainly behind paywalls so using school paper, which seems to have a lot of good student journalism.
It seems like the Provost has been not reporting his use of LLMs appropriately to academic journals, and in one case, used AI entirely to write an op-ed in a paper that doesn’t allow AI at all.
He also was trying hard to hide his use…
According to the committee, Schnell uploaded the manuscript into Dartmouth Claude enterprise and asked, “Are there many signatures of AI writing in this article?” The committee wrote that Schnell later submitted the manuscript, incorporating many of the changes that Claude suggested in the article.
Personally, I am not ant-LLMs/AI for all purposes, at all, but this is not ok. I am glad they held a PROVOST to account.
“The Provost used AI to meaningfully alter his writing in violation of at least one publication’s AI use and/or disclosure requirement, failed to disclose his use of AI in clear violation of a national newspaper’s stated policy, was not forthcoming about his AI use, and fell short of the standards set forth by Dartmouth’s Code of Ethical Conduct, the Academic Honor Principle, and Geisel’s Policy on Expectations for Professionalism for Faculty, Non-Faculty Academics, and Staff,” the committee concluded.
Schnell was accused of using AI to alter his work by the student newspaper The Dartmouth and the outlet Semafor, specifically pointing to an opinion piece the provost wrote for The Washington Post.
If you read the school newspaper piece, which digs into the investigation, it is pretty clear where Provost went awry.
He lied about his use and/or used it w/ publications that didn’t allow AI at all. It actually wasn’t that complicated in my view. Very old-school ethics in the end! Especially ironic given the provost, is, apparently very Catholic.
I should add, if one is an academic, it may be complicated keeping tabs of which journals have which rules. That may be tricky, but I do think that is part of one’s job.
And instead he apparently asked Claude how to make his rule violations harder to detect.
That is not innocent behavior. And it was naive, because AI isn’t good at letting you know when what you are asking for is not actually a good idea, including because it probably won’t work.
He could have asked Claude to summarize the rules. That’s what I meant. It is not that complicated, folks. The permissible use of AI is usually clearly spelled out. Here is, for example, the guidance for the ACS publications, as summarized by Google:
Sorry, to clarify, I understood your suggestion and why it is ethical and appropriate. I was then pointing out he did something very different, and not ethical.
It seems to me that a high-functioning adult (certainly a skilled PhD who has risen to the level of Provost) can write circles in both grammar and idea generation compared to a Large Language Model.
Has AI really improved so much in the past year or so that I am underestimating it?
He was using it for coding stuff as part of his academic papers, in part and yes it is incredibly useful there by all accounts. I a not a chemist so don’t entirely understand advanced chem coding though;).. however all my tech friends have told me it has entirely change their jobs.. like it is night and day from 5 years ago. I think those in other industries don’t understand.
Also there are also models you don’t have access to as a “regular” person for free (not saying that is what he used to write an op-ed, probably not) BUT if you use LLMs randomly here/there you likely have no idea the power of them for someone who is expert at their use with prompt engineering and using GPTs and all the bells and whistles.. using GPT like a search engine is nothing like what experts do. If you train it, you can make them sound like you, not like “it”. You basically build tools that churn out papers/op-eds, grants that use your own data/source materials to sound like your own writing. This takes time to build and some skill to build on front end, though.
About a month ago, a somewhat-relative-by-marriage published an article in a well known publication about AI.
He is a very successful professional who lectures at UT-Austin’s business school.
The published article was about 12 paragraphs long. The first 8 paragraphs were clearly written/assisted by AI. The word choice was incredibly accurate, varied, and appropriate for the first two-thirds of the article. The last 4 paragraphs were not as well written and contradicted the main theme of the first 8 paragraphs. The difference was obvious.
