Wikipedia:Bots/Requests for approval/Zentralbot

The following discussion is an archived debate. Please do not modify it. To request review of this BRFA, please start a new section at Wikipedia:Bots/Noticeboard. The result of the discussion was

Denied.

Zentralbot

New to bots on Wikipedia? Read these primers!

Approval process – How this discussion works
Overview/Policy – What bots are/What they can (or can't) do
Dictionary – Explains bot-related jargon

Operator: Physikerwelt (talk · contribs · SUL · edit count · logs · page moves · block log · rights log · ANI search) PhilMINT (talk · contribs · SUL · edit count · logs · page moves · block log · rights log · ANI search)

Time filed: 17:52, Wednesday, December 2, 2020 (UTC)

Function overview:

Supervised

Programming language(s): Python

Source code available: https://github.com/ag-gipp/AnnoMathTeX/tree/b29acb02834b41ab7b989a82c787e689129e0194/evaluation/wikipedia-export

Links to relevant discussions (where appropriate):

Edit period(s): one time

Estimated number of pages affected: about 30 See https://github.com/ag-gipp/dataAnnoMathTex/tree/master/evaluation

Namespace(s):0

Exclusion compliant (Yes/No):

Function details: See https://github.com/ag-gipp/AnnoMathTeX/issues/11

Discussion

Hmm. Bots aren't usually approved for one-time low-volume tasks; if it's just a 30 page one-time run have you considered just doing this manually? For example, if you use your script to get the new content of the article (after modifications), then (with that output) update the content on Wikipedia using your main (current) account yourself, that'll be fine. ProcrastinatingReader (talk) 18:31, 2 December 2020 (UTC)[reply]

You are right. @PhilMINT: could even use an oauth token associated with his account. Since he decided on each individual change manually, this would even better represent his interlectual effort. I am happy with this option. However, I see two arguments for a bot account. 1) If for any reason anything goes wrong one can nuke the contributions easily. 2) I wanted to learn how the bot process works for future tasks, like phab:T268935 given the complexity of the last math bot phab:T197925 see (http://phabricator.wikimedia.org/diffusion/TTEX/browse/master/mathwikibot.py). Physikerwelt (talk) 04:07, 3 December 2020 (UTC)[reply]
- You're editing existing pages right, rather than creating new ones, so I don't think it would need nuking ever? (there are additional limitations on article creation by bots if creating as well).
  The bot approval process is not too scary when you're ready for it :) -- basically it's just a period of time to allow for comments, evaluating whether the task can & should be done, and then a technical trial to make sure it isn't messing up. But since this process takes a bit of time, and we don't have too many active approvers either, low-volume one-time 'bots' are usually advised to edit manually instead, so the approval process mainly deals with regular tasks / high-volume one-time tasks. Bots are only needed when you're not reviewing your changes. As long as you review each change before you hit "save" (either by hand, or the program showing you a diff that you confirm is OK before save) a bot account is not needed: this is called "semi-automated editing" (see WP:BOTASSIST).
  If, however, it's a lot more work (with your given codebase) to output the change for manual editing (to the point where it's likely the task won't be done), do say. Obviously it's better to do a trial if the alternative is nothing will be done at all. ProcrastinatingReader (talk) 08:42, 3 December 2020 (UTC)[reply]
- I have finished implementing the bot and tested it in my sandbox (which caused some side-effects for semi-protected pages). See https://github.com/ag-gipp/AnnoMathTeX/pull/12. I will double check with @PhilMINT: if the changes represent his intention. If so, the changes can be performed under his username without violating complience. Is that correct? Physikerwelt (talk) 15:12, 3 December 2020 (UTC)[reply]
  I've looked at your pull request very quickly. It seems like you've made it so it will edit a new copy of the article in the sandbox, and then you intend to copy it over to the live article by hand (using your own account, not the bot one)? If so then yes, that looks okay to me. ProcrastinatingReader (talk) 16:03, 3 December 2020 (UTC)[reply]

User PhilMINT ran this script/bot on 2020-12-09 on some 25 articles, which – in my opinion – introduced many inappropriate qid links: at least half of them lead to equations that have no resemblance to the linked ones. A few of these are discussed at User_talk:PhilMINT, but there are many more. It'd take a lot of manpower to check their accuracy and correct those that are wrong. I am not sure what value these links add to wiki articles (they surely add more machine code to wiki markup) even if they lead to apparently correct pages but with trivial content. My question is: was the script approved for such use, and what should be done to revert the edits en masse? It's better to not have qids than to have the wrong ones, IMO. Ponor (talk) 16:56, 10 December 2020 (UTC)[reply]
It's not "approved" in the sense that this is semi-automated editing. The person making the edit should manually check it is correct, and if so they can click save. Such an edit is treat exactly the same as any other edit made on the encyclopaedia. In that sense, it shouldn't be problematic (because the person has checked it before saving), but other editors are free to revert if they disagree with the edit, and then discuss on talk pages, in line with the normal dispute resolution process / WP:BRD, etc. ProcrastinatingReader (talk) 17:02, 10 December 2020 (UTC)[reply]
Given so many errors, some of which are very obvious, I don't think the changes were manually checked. The thing with such scripts is that the damage they make in 4 minutes may take a few hours to fix. I'd rather do something more useful here than chasing hundreds of edits by a script. Hopefully an admin with the right tools will show up to help. Ponor (talk) 17:14, 10 December 2020 (UTC)[reply]
Apparently it's ~20. You can tap the "undo" button on them all if you disagree with the edits; should take about a minute. ProcrastinatingReader (talk) 17:17, 10 December 2020 (UTC)[reply]
That being said, I've checked the one at Velocity and it doesn't seem like it caused errors, even if one may disagree on if the statements should be clickable as a matter of content. ProcrastinatingReader (talk) 17:22, 10 December 2020 (UTC)[reply]

Looks like we, the editors, are the main subject of this experiment. Not sure that's ethical (considering waste of time), but won't go into that. Yes, some links are OK. Most likely not needed but OK. In the articles I checked, I'd say some 50% were either wrong or pointed to pages with trivial contents, thus unverifiable. To check and correct all qids added in all 20+ articles (20×5=100) would definitely take more than a minute, unless I assume those edits were made in bad faith and revert them without checking at all. I am glad this was a one-shot thing, I feared it's something much worse. Once more edits are added to the affected articles, these won't be that easy to revert. Ponor (talk) 17:59, 10 December 2020 (UTC)[reply]

The above discussion is preserved as an archive of the debate. Please do not modify it. To request review of this BRFA, please start a new section at Wikipedia:Bots/Noticeboard.