GNU social JP
  • FAQ
  • Login
GNU social JPは日本のGNU socialサーバーです。
Usage/ToS/admin/test/Pleroma FE
  • Public

    • Public
    • Network
    • Groups
    • Featured
    • Popular
    • People

Conversation

Notices

  1. Embed this notice
    Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 29-Jun-2023 06:04:26 JST Paul Cantrell Paul Cantrell
    • Carl T. Bergstrom

    Louder for the folks in the back!

    “Many (all?) current ChatGPT detectors have not been adequately assessed for issues of algorithmic bias and therefore should not be used to accuse students of misconduct in their written work.”

    All carefully stated, with appropriately judicious academic language, but these are some pretty stark “unsafe at any speed” results.

    Do not use putative AI detectors in your classroom. Just don’t. Please

    From @ct_bergstrom:
    https://fediscience.org/@ct_bergstrom/110623757885080991

    In conversation Thursday, 29-Jun-2023 06:04:26 JST from hachyderm.io permalink

    Attachments


    1. Domain not in remote thumbnail source whitelist: fediscience.org
      Carl T. Bergstrom (@ct_bergstrom@fediscience.org)
      from Carl T. Bergstrom
      Ironically, these false positives are readily avoided by asking ChatGPT to rewrite the non-native English speaker's text to increase linguistic complexity. In other words, the way for these speakers to avoid being accused to cheating is to actually cheat. The take-home for higher ed is obvious and stark. Many (all?) current ChatGPT detectors have not been adequately assessed for issues of algorithmic bias and therefore should not be used to accuse students of misconduct in their written work.
    • Embed this notice
      Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 29-Jun-2023 06:07:58 JST Paul Cantrell Paul Cantrell
      in reply to

      I’ll add an incautious, injudiciously stated Paul note to the above:

      The problem isn’t merely that the products don’t work. The problem is that these products are trying to solve — automating the detection of cheating — is a nonsense problem.

      If you set out looking to oil snakes, •you will end up buying snake oil•.

      In conversation Thursday, 29-Jun-2023 06:07:58 JST permalink
    • Embed this notice
      Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 29-Jun-2023 06:11:26 JST Paul Cantrell Paul Cantrell
      in reply to

      improving pedagogy > policing pedagogy

      In conversation Thursday, 29-Jun-2023 06:11:26 JST permalink
    • Embed this notice
      Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 29-Jun-2023 06:34:52 JST Paul Cantrell Paul Cantrell
      in reply to
      • Carl T. Bergstrom
      • Cleopatra

      @Cleopatra @ct_bergstrom
      Oh, the “real” exams have even more problems: they still allow cheating, they have accessibility problems up the wazoo, and they’re absolute •shit• for retention. It turns out that cramming for an exam is a terrible way to learn things, big shocker.

      Mass education is a tricky problem no matter how you slice it. Educating and evaluating at the same time multiplies the problems tenfold. And then when you try to increate the student-teacher ratio…

      In conversation Thursday, 29-Jun-2023 06:34:52 JST permalink
    • Embed this notice
      cleopatra@c.im's status on Thursday, 29-Jun-2023 06:34:53 JST Cleopatra Cleopatra
      in reply to
      • Carl T. Bergstrom

      @inthehands @ct_bergstrom
      That looks like it can get pretty nasty ☹️.
      Cheating-detection is the kind of stuff where you have to be *sure*, and if your algorithm just searches for people writing in simple and broken english then it won't be the rich white guys getting accused.

      imo there's a related issue that was already exposed by 'the shadow scholar' a few years ago: too many damn essays, too few real exams where you have to sit in a room and where cheating is pretty hard.

      In conversation Thursday, 29-Jun-2023 06:34:53 JST permalink

Feeds

  • Activity Streams
  • RSS 2.0
  • Atom
  • Help
  • About
  • FAQ
  • TOS
  • Privacy
  • Source
  • Version
  • Contact

GNU social JP is a social network, courtesy of GNU social JP管理人. It runs on GNU social, version 2.0.2-dev, available under the GNU Affero General Public License.

Creative Commons Attribution 3.0 All GNU social JP content and data are available under the Creative Commons Attribution 3.0 license.