My syllabus statement on AI generative text. Feel free to reuse CC0 and/or critique here with responses. Responses also welcome if you’re gonna use it (I might put a tally in the annual report)
@josh@academicchatter I think it's entirely reasonable. You're allowing people to use the tools as they see fit, providing the guidance that you don't think they will be useful and making clear that you need to see what their own contribution is. I can't think of anything that it lacks.
@josh@academicchatter I think this is an excellent statement. I especially love the citation format. This was a debate point at our institution (that there is no established citation format).
@josh@academicchatter
If I were one of your students and read this, I'd be highly skeptical that AI generated text would be graded the same as other text, whether consciously or unconsciously. Your preconceptions, which granted may be based on previous experience, are clear.
It would be interesting to see blind grading without knowing what AI wrote vs. not.
@ViennaMike@josh@academicchatter I lean pessimistic so I think this will be ineffective. Students generally wait till last minute to write and thus get desperate. LLM AI output will always be better than nothing. Since you can never prove it was written by AI, there's no need for them to follow your instructions.
@rspfau@ViennaMike@academicchatter
My writing assignments require lots of citation use w/course materials & other sources students find on their own w/univ library search engines. I'm allowing AI as an added tool if they wish, with warning that it's not v good. LLMs don't do complex references or reviews. I tested my assignments & feel v comfortable.
@GetzlerChem@academicchatter
If there’s anything to report. Last year I had them work together to compile and synthesize their individual results from research papers about Mastodon. Then they tried to make ChatGPT reproduce equivalent quality information. It failed badly. It seems to me the vital factor is references and citation. A llm can’t muster and organize complex and specific referents, but statistically summarizes sentiments.
Not a critique of the text per se, but one should distinguish between essays as a learning tool from essays as a validation method (exam).
The latter has been flawed well before generative text, due to the inherent difficulty of detecting (and defining) plagiarism. Time to ditch the whole concept, in my opinion.
Some thoughts. What if the text is blatant plagiarism of a real author? Like, ‘I had no idea that paragraph was Malcolm Gladwell / Jordan Peterson / Taylor Swift, the computer just said …’
Also, do you have any assignments, like an in class final, where this wouldn’t be an option? If so, you might want to clarify that AND point out that the purpose of the assignments is to learn, so even if the homeworks get passing grades, there may be major gaps in learning.
@JMMaok
Good point. If there’s some failure of data memorization that results in effective plagiarism and I catch it, then that’s “unacceptable” per my conditions. I’m not really worried about it. My requirements include lots of specific reference use that can’t be replicated with AI. If I’m wrong (unlikely) then I learn about weaknesses in my requirements.
Add comment