Using Artificial Intelligence to Create OER


While I think it is a nightmarishly bad idea for the open education world to use generative AInt for OER … I don’t know that I think it is a copyright problem, in many of the situations in which it is used.

Warning: IAmNotALawyer, and this is not legal advice.

But my understanding is that the copyright situation regarding the outputs of generative AInt falls into two categories, based on where the use is made: outside of the US, I think there are as yet very few formal legal regulations or laws or decisions — so, maybe, anything goes?  (The EU is working on its „AI Act“, but that actually has very little to say about issues like the status of genAInt outputs.)

So to the extent that this ruling applies to OER, the entire work will be born into the public domain, or else there will be a thin copyright if an OER author selects and arranges public domain genAInt outputs, or else there will be a large, human-created, OER with author as rightsholder into which the creator will have inserted a few public domain genAInt outputs — in the latter two situations, the human will have a copyright they can use to apply a CC license, while in first situation, the whole work is in the public domain … which counts as OER (by the UNESCO OER Recommendation definition of OER, which I usually use as the canonical definition of OER … but most other definitions agree on this point)!

The nightmare of genAInt in OER is instead because it is basically a fancy autocomplete tool, why in the world would we create educational materials with fancy autocomplete?

Even if one could find an answer to that question („why would we…“), there are so many downsides to genAInt:

  • it generates reasonable-looking text which can be filled with errors
  • it has no understanding of anything, so cannot write with insight, empathy, pedagogical experience…,
  • it is often based on exploitative labor (e.g., LLM companies are literally outsourcing the content moderation which is supposed to prevent horrific (even illegal, such as CSAM) material being generated to terribly paid workers in developing countries),
  • it is based on material scraped from the ’net which could be a copyright violation itself (legal experts disagree; personally, I do think there are copyright violations in making and distributing the models), and which reiterates the biases and injustices which are so rampant on the ’net
  • it consumes enormous amounts of power and water (like the last tech hype cycle to go bust recently, the one based on blockchains, the server farms for AInt are consuming power which approaches that of an entire country!)
  • it sends exactly the opposite message we are trying to send in education, that we care about people and we believe there is a value in people knowing things and gaining expertise … rather we want fancy autocomplete to do the work of human experts,
  • and on and on….

Oh, I guess there are two responses I can imagine people giving to my „why would we hand education over to fancy autocomplete?“ question:

  1. It will be used by experts to get a start, they will then edit the genAInt output and fix all the errors.  This will save them time and effort.
  2. We don’t have the money to pay for people to do the important work of making OER, so this is a way to get something … and something, even of very low quality, is better than nothing.

To which I would reply:

First of all, there are all of those big, systemic problems mentioned above, so these lukewarm justifications are doing a lot of work!

Second, for point 1: Well, OK, if it saves you time, go for it.  When I make videos, I use YouTube’s voice recognition autocaptioning … and then spend a lot of time fixing all of its mistakes — but I think real experts will have to spend a very large amount of time and energy to get genAInt OER into usable form, putting in all of the pedagogical insight and correcting all of the (plausibly worded) mistakes genAInt makes.  I would be very surprised if this were a faster way to make higher quality OER.  …I guess, up to the system problems already mentioned, I’d be happy to be proven wrong about that.

Finally, for point 2: if it’s just a question of money and institutional priorities … I’m willing to keep fighting the good fight against the defunding of quality education, I don’t see why open education folks accept this more draconian attack on quality education and on educators‘ livelihoods — other creatives fought against this nonsense (see: the Hollywood writers‘ strike), why aren’t we?  Instead, we seem to be happy to grease the slide into this particular dystopia.

Sorry, that got a bit rant-y.  I apologize.  My main point was that I don’t think copyright will save us here.



Title: Using Artificial Intelligence to Create OER
Source: Stephen’s Web ~ OLDaily
Source URL:
Date: October 25, 2023 at 11:40PM
Feedly Board(s): Schule,Englisch