Re:DARPA Moves Forward on X-65: Plane with no moving control...
by lamontcg on Tuesday January 09, 2024 @08:12AM
(#38925762)
It might be useful to have LLMs do automoderation on forums, tweaked by the moderators to shift the discussion the direction they want it to go. Have it so that the feedback is immediately on pressing the submit button rather than after-the-fact. And probably with more useful information to the poster than the "our users don't post messages like this" message.
I think it'd be useful to ban pun threads and other low effort shitposting, but it would probably have issues determining actual expertise from faux expertise. With the wrong settings it would be frustratingly useless, but over time some useful defaults should emerge.
There's a spectrum between Reddit on one end, and the typical IRC channel fiefdom with its mod-permissioned voicing, invisible banhammer tripwires, and admonitions to LuRK MoAR, on the other, and I think HN is in a sweet spot/zone somewhere between those extremes, but perhaps might need an adjustment in the direction of the latter.
Downvoting/flagging somewhat furnishes this signalling, but it's too ambiguous of a signal and is often abused by people who simply want to suppress information that is perfectly valid but not to their liking.
This is where a slashdot type of lightly sentiment-coded moderation system really may be worth bringing back into fashion.