Originally published at: Transparency of Generative AI Use in OEAward Nominations – OEGlobal
In each passing day of reading online content, an ongoing question in the back of my mind is, “was this written with Generative AI?” Being able to confidently detect synthetic text is debatable, thus we are seeing much interest in how content creators are being transparent in stating their GenAI use. Recognizing the expanding use…
3 Likes
Here’s a little something from a reviewer (me) on how I used AI for my reviews. (This is the copy from an email I sent to Alan, originally.)
A new EU regulation is about to take effect requiring content created with AI to be labelled. I am curious to see how that plays out, because much of the public debate misses the point: very little content of any relevance is simply “made by AI”. Most of it emerges from a ping-pong between human and machine. That is how I worked here too. In short: I read each nomination and annotated it directly, with comments, highlights and emojis. I gave an AI the logic behind the award forms so that it could work within that framework. For each nomination I then recorded a half-structured voice note in German, talking through my notes and emojis — sometimes freely, sometimes along the questions in the form. The AI transcribed and translated this, converted my unordered thoughts into the logic of the criteria, and produced a first draft. It then asked me questions about what was still missing, unclear or contradictory in my assessment, and produced a second draft from my answers, which I transferred into the form. My conclusion: the judgement is overwhelmingly mine, the wording is largely the AI’s.
When I read AI-generated texts in German, my native language, I often find them quite irritating — especially when reading closely, they tend to sound empty and hollow. So when I write in German, the actual wording is never something I would hand over to an AI tool. It’s different when I write in English, because it isn’t my native language and I don’t feel confident writing in it. Here I often use translation tools, or I even just feed my key points into an LLM and have it phrase them in English for me.
I’d be curious to know how such texts come across to native English speakers (or to people with a very good command of English). Do they sound as hollow and empty as German AI-generated texts do to me? Or do the ideas and the words that are in my head in German still find their way into the English text?
That gave me a spontaneous idea: it might be a good thing — perhaps for next year’s awards — to offer English-speaking mentors you could meet online, explain your idea for a nomination to in broken English, so that the enthusiasm hopefully carries across, and then get support in actually writing the text. Not in the generic AI register that a translation tool or an LLM would produce, but in a way where the real intention is picked up by an actual human being.
(I wrote this text in German as well and then had Mistral translate it and turn it into “proper” English. To me it reads coherently in English, but I don’t know whether what I’m really getting at still comes through.)
Your translated post here Nele doesn’t read as GenAI-ish hollow, it does read to me as conversational. I’m far from expert but I had understood from readings of linguist Emily Bender that translation is more machine learning than AI.
I am not necessarily sure any more that we can make an AI or not call, so to me the best route is being transparent.
As far as the awards, indeed maybe we could try to organize a way to get feedback. There is no requirement that nomination be written in English, but that seems to be an assumption. We had a few this year not many in other languages.
It might be a matter more of how nominators think they have to use academic level writing. I do get to scan them all, but am not an evaluator, but I would find a more human written nomination that is persuasive on the relation to the ideas of open education is more important than citations and references and sounding lofty (this is my opinion). I do think maybe it’s an issue of “Excellence” in the name of the awards, I’d like to have it focus more on recognition.
Translation tools I use almost daily (like to read your Mastodon and blog posts). The approach Jöran describes above and your own description for helping formulate ideas for direct communication all seem reasonable.
Thanks for replying here (and congratulations on you nomination recognition).