Conversation

Notices

1. Embed this notice
   Adam Shostack :donor: :rebelverified: (adamshostack@infosec.exchange)'s status on Monday, 29-May-2023 02:49:48 JST Adam Shostack :donor: :rebelverified:

   I could swear I saw a post here about a professor who had their students grade an assignment from ChatGPT, but Masto search is a dumpster fire.

   If you saw it and can share a link I'd be grateful.

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Monday, 29-May-2023 02:49:40 JST Paul Cantrell

     in reply to

     • Steve Bellovin

     @adamshostack @SteveBellovin
     Brain dump, in the hopes that some bits help:

     Group projects are •fantastic•…if they work. But they are much harder to scaffold and manage. I recommend jumping into them if and only if you are ready to undertake the effort it will require on your end. (Can say more if you want.) Absent that work, you’ll end up with muddling, mooching, and resentment. This work is to a large extent a function of who your students are, and what experiences they bring working on teams.

   • Embed this notice
     Adam Shostack :donor: :rebelverified: (adamshostack@infosec.exchange)'s status on Monday, 29-May-2023 02:49:41 JST Adam Shostack :donor: :rebelverified:

     in reply to

     • Paul Cantrell
     • Steve Bellovin

     @SteveBellovin @inthehands I'd seen that -- TBH, I have a hope that my students will do their own final projects. Someone else privately suggested making those projects into group projects to create social pressure to do the work.

   • Embed this notice
     Steve Bellovin (stevebellovin@mastodon.lawprofs.org)'s status on Monday, 29-May-2023 02:49:44 JST Steve Bellovin

     in reply to

     • Paul Cantrell

     @adamshostack @inthehands Right. I had posted https://mastodon.lawprofs.org/@SteveBellovin/110419015739917805; one of the responses is the thread I replied to you with.

   • Embed this notice
     Adam Shostack :donor: :rebelverified: (adamshostack@infosec.exchange)'s status on Monday, 29-May-2023 02:49:45 JST Adam Shostack :donor: :rebelverified:

     in reply to

     • Paul Cantrell
     • Steve Bellovin

     @SteveBellovin Thanks! My current plan is to allow use of LLMs, but you have to submit your prompt, the LLM's response, and your edits to it as a change-tracked word doc.

     I'm working on a grading rubric that rewards varying your prompt, getting multiple answers, and maybe more.

     (cc @inthehands )

   • Embed this notice
     Steve Bellovin (stevebellovin@mastodon.lawprofs.org)'s status on Monday, 29-May-2023 02:49:47 JST Steve Bellovin

     in reply to

     @adamshostack https://hachyderm.io/@inthehands/109479808455388578

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Monday, 29-May-2023 02:52:56 JST Paul Cantrell

     in reply to

     • Steve Bellovin

     @adamshostack @SteveBellovin
     Having students share how they used AI output is promising. In general, opening up the writing process to allow instructor-student communication throughout (not just after handing in) is a powerful strategy.

     If it were my class, I’d put the emphasis on reflection and critique of GPT output, not just on mechanics of sharing. Ask them to react, study their reactions, reflect on them, be mindful about all of that. Maybe have them hand that in •before• finished paper…

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Monday, 29-May-2023 02:54:39 JST Paul Cantrell

     in reply to

     • Steve Bellovin

     @adamshostack @SteveBellovin
     …and give just the quickest feedback on it. Tiny thoughts in passing during the writing/revision process go a long way!

     Related: emphasizing the idea that GPT output is generally (1) full of incorrect information and (2) bad writing both helps discourage cheating and opens up those channels of critical thinking.