Timeline for answer to GPT on the platform: Data, actions, and outcomes by Chris
Current License: CC BY-SA 4.0
Post Revisions
14 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 8, 2023 at 11:30 | comment | added | Martijn Pieters | There have also been a few escalations of plagiarism suspensions, and again the decisions were upheld. I've seen the CMs re-verify the evidence for specific posts each time, and also initiate talks with specific tech companies whose freelance support teams have been plagiarising widely in the name of "providing tech support" via SO. But, for ChatGPT, the picture is different, and without having experienced what John Ericson calls the "Barnum Effect" can be bewildering and hard. The evidence is not easily displayed as metrics on a screen. | |
| Jun 8, 2023 at 11:26 | comment | added | Martijn Pieters | I can actually see why the base rate would have been 0 before. The majority of suspensions are for voting fraud or CoC breaches (being abusive, basically). I can imagine that the majority of appeals would be for voting fraud issues, because users have no visibility on how much evidence we have for voting fraud. I've actually been asked by CMs to share any further evidence in a few such cases, and they invariably are from people that think that they had covered their tracks sufficiently to have gotten away with it. In all those cases I was involved with the moderator decision was upheld. | |
| Jun 8, 2023 at 4:55 | comment | added | Alex | The appealer might present evidence to support their appeal. Or the moderator’s action might have been unambiguously wrong. | |
| Jun 8, 2023 at 4:42 | comment | added | Chris | @Alex Why else would they overturn a suspension? Because the evidence was too good? | |
| Jun 8, 2023 at 4:37 | comment | added | Alex | Either way, in context, I think his point is the same. The difference between a regular appeal and a ChatGPT appeal is that in the former, the reviewer can see the evidence while in the latter the reviewer cannot. | |
| Jun 8, 2023 at 4:34 | comment | added | Alex | Though actually reading it again, I’m not sure if the insufficient evidence is referring to the moderator or the appealer. | |
| Jun 8, 2023 at 4:32 | comment | added | Alex | “rare and notable” is part of the same sentence. It doesn’t say that it’s rare and notable to overturn a moderator’s decision. It says that it’s rare and notable to overturn a decision due to insufficient or contradictory evidence. | |
| Jun 8, 2023 at 4:30 | comment | added | BryKKan | Have I just been telling them to spit into the wind all this time? Yeah, probably. I can't really imagine that there wouldn't be at least an occasional case where you changed your mind upon seeing new information. I would expect this also extends to reevaluating the quality or applicability of other evidence within a new context. A fresh set of eyes, looking sincerely and impartially, ought to discover such cases even more often. I don't see how a zero reversal rate, even if just "0 for insufficient evidence", is anything other than an admission they aren't taking all appeals seriously. | |
| Jun 8, 2023 at 4:00 | comment | added | Chris | @Alex You skipped "rare and notable." And "no evidence on either side" is ridiculous when the CMs haven't even discussed the suspensions in questions with the moderators. | |
| Jun 8, 2023 at 3:41 | comment | added | Alex | "due to insufficient or contradictory evidence" I take that to mean that it is possible in general to present sufficient evidence that a moderator erred (or acted maliciously), and then the Community Managers can overrule it. The issue with ChatGPT appeals is that there is no evidence on either side, so there is no way for the Community Managers to review a decision. | |
| Jun 8, 2023 at 2:41 | comment | added | Chris | @Alex "It is rare and notable if we are ever in the position of overturning a moderator’s decision due to insufficient or contradictory evidence." Seems pretty clear to me. And now they have a whopping two dozen cases and it's too much? Give me a break. At least talk to the moderators in question before you freak out. | |
| Jun 8, 2023 at 2:22 | comment | added | Alex | Was the post saying that they never overturn moderator decisions on appeal? I read it as saying that they never have to overturn it due to being unable to verify something. I.e. there may well be plenty of times where they can review a moderator's actions in other areas and declare that the moderator was wrong, but with ChatGPT appeals they simply have no evidence to work with. | |
| Jun 7, 2023 at 21:04 | history | edited | Chris | CC BY-SA 4.0 |
added 290 characters in body
|
| Jun 7, 2023 at 20:47 | history | answered | Chris | CC BY-SA 4.0 |