Faculty don't just know things. They do things. This is why we should be thinking of the labor of teaching and research, not just the products and outcomes. - ThreadSky

biblioracle.bsky.social • 77 days ago

Faculty don't just know things. They do things. This is why we should be thinking of the labor of teaching and research, not just the products and outcomes.

Reposted from Hollis Robbins

Faculty must make the case that their knowledge is outside of AI, that a professor knows something AI does not yet know. Otherwise why bother?

Comments

kasenc.bsky.social•77 days ago

I just want to say I really appreciate you and your perspective and patience. I hope you have a nice day today.

freyaanduin.bsky.social•77 days ago

once wrote about what I use that cannot be found with AI

https://lendorphandlacour.dk/outside-ai/

badpacifist.bsky.social•77 days ago

There's another distinction worth drawing here, between "conveying information" and "teaching." Conveying information is just putting it on offer. Teaching actually involves attention to the human processes of learning and the relationships that enable those, which AI cannot do.

biblioracle.bsky.social•77 days ago

Absolutely. Teaching and learning is an experience, an exchange, not just a one-way conveyance, and also not something that looks the same for everyone every time.

wesflinn.bsky.social•77 days ago

I will say something intemperate that may cost me current and future employment:

At most schools, even those which claim to be teaching-focused, research is weighted too heavily in the review/tenure/promotion process. This is because we almost all trained at R1s.

wesflinn.bsky.social•77 days ago

And yes, that doesn’t sound directly related to the topic at hand, but it also explains why certain faculty are happy to outsource teaching.

badpacifist.bsky.social•77 days ago

OK, now I can see. You could skip past the first 5-6 pages of that article and get to the guts of it, which largely don't get stuck on disciplinary boundaries.....

badpacifist.bsky.social•77 days ago

At the risk of self-promotion, I take up that issue in my field here (didn't look to see if I could see yours without cyber-stalking!). And if anyone threatens your job prospects because you said this, please send them to me. Or me to them. Whichever.

wesflinn.bsky.social•77 days ago

I am sharing this article with colleagues right now.

badpacifist.bsky.social•77 days ago

I hope it makes dents in some bad habits we have. I've been in a 4/4 teaching job for 23 years and only recently have people stopped asking me when I was going to leave for a "real job" somewhere so I could publish more. Your point is the mirror image of that.

badpacifist.bsky.social•77 days ago

Unfortunately, technocracy is all about erasing, at least backgrounding, human relationships (I know you know this--you've been saying it for years). I'll never understand why that doesn't bother some people more, especially if they're not direct beneficiaries of the systems that result.

tkedmunds.bsky.social•77 days ago

True story. We create the knowledge that gets fed into the AI grist mill.

anecdotal.bsky.social•77 days ago

Agreed -- what faculty know and do are things outside of what AI knows. (I'm speaking of AI as the "knowledge commons" broadly.) But I argue that the labor of teaching someone what AI knows is not really labor that anyone should pay for.

biblioracle.bsky.social•77 days ago

This same argument could be applied by substituting Wikipedia for AI. Both are examples of the "knowledge commons." You're ignoring what's meaningful about the experience and activity of teaching and learning and elevating AI into something it isn't.

anecdotal.bsky.social•77 days ago

Which AI platform do you use and how do you use it? This is a serious question, not trolling. I use Claude 3.5. I know what it does not know about my field and I know what it doesn't know.

biblioracle.bsky.social•77 days ago

I experimented pretty extensively with half a dozen or more (both general and more purpose built) models as I wrote my forthcoming book and found little ongoing utility for my work because I'm a writer and my writing requires MY thinking about my subjects, not relying on a statistical...

biblioracle.bsky.social•77 days ago

...representation of what's in a model's training data. In my experimenting, I had thought I would find some utility to my work, but over time, I found them more hindrance than help as skipping parts of the process by using an LLM required me to double back and start over.

biblioracle.bsky.social•77 days ago

There's also the core problem of what a model "does" or "doesn't" know may shift from moment to moment because of the problem of hallucinations. I don't know how to integrate the work of something that tells me highly plausible, but untrue things.

palan57.bsky.social•77 days ago

AI does not actually know things in any conventional sense of the word.

Comments

Posting Rules

Reply