Discourse AI silence link takes you to an inaccessible link as a normal user

,

This is relatively self-explanatory. The silence link takes you to an admin link. Not sure this is fixable, but I do not think this link can be accessed.

2 לייקים

I was thinking about bringing this up few months ago. It’s the link given by Discourse AI when silencing users, but I don’t think it’s a bug :thinking: . The text can be changed here:

I recommend changing this to link to the profile of the machine account that performs the silencing. The site administrator can provide any additional information they would like in that account’s profile.

I did notice that I was not able to see an account with the default username discourse_ai_spam on this forum. Maybe that is only because the username of that machine account has been changed to something different on this forum.


If linking to the machine account profile page is considered undesirable for some reason, or not feasible, the alternative would be linking to this URL: https://www.discourse.org/ai

But can’t that already be done by clicking on the avatar in the user card?

No, that will take you to the profile of the silenced user.

There are two separate accounts involved:

  • The account of the user who was silenced due to their post being detected as spam
  • The machine account that performed the silencing operation

My proposal is that the link point to the profile of the machine account that performs the silencing.

Whoops, my apologies - I misread your post.

לייק 1

I think last year, the review queue had some checks so it showed the same text to admins, moderators, and category moderators, but there was only a link for those who were able to use it. So there were no links for category moderators who cannot access the admin area, some links for moderators (for example, on watched words), and plugins like Askismet had a link only for admins. (I think AI was a link for everyone there too, so maybe I have a screenshot of that; I will check later.)

But my point is maybe the same solution makes sense here: Only provide a link for those who can see the AI configuration.

2 לייקים

I think it is important to clearly communicate the nature of the action to everyone. The reason is that, although it is a very useful tool, the AI spam detector is prone to false positives.

There is a significant difference between the implications of an automated system silencing a user as compared to a human moderator doing the same. If people assume the silencing was performed by the human moderator, then that unfairly attaches a stigma to that new user in the case of a false positive.