Wikipedia:Bots/Requests for approval/Cewbot 4
- The following discussion is an archived debate. Please do not modify it. To request review of this BRFA, please start a new section at WT:BRFA. The result of the discussion was Approved.
Operator: Kanashimi (talk · contribs · SUL · edit count · logs · page moves · block log · rights log · ANI search)
Time filed: 11:29, Wednesday, February 26, 2020 (UTC)
Function overview: To inform of the article's PROD eligibility
Automatic, Supervised, or Manual: Automatic
Programming language(s): wikiapi on GitHub
Source code available: 20200206.reminded expired AfD.js on GitHub
Links to relevant discussions (where appropriate): Wikipedia:Bot_requests#A heads up for AfD closers re: PROD eligibility when approaching NOQUORUM
Edit period(s): daily
Estimated number of pages affected: ~20/day
Namespace(s): Wikipedia
Exclusion compliant (Yes/No): Yes
Function details: Get infomations from previous PRODs, previous AfD discussions, previous undeletions, and current redirect. Sample report.
Discussion
editSpeaking as a regular AFD closer, I don't think the bot needs to say whether there is a quorum or not. I realize that this slightly deviates from the purpose of the bot, but is probably more useful it just says that there was a previous PROD, undeletion, AFD etc. Jo-Jo Eumerus (talk) 11:39, 26 February 2020 (UTC)[reply]
- @Jo-Jo Eumerus: Thank you for your comment. The note "There are participations and the report will not shown in the deployment environment", as it says, will not shown in the AfD. It is only a log. But I wonder if I can use the information to do more things... --Kanashimi (talk) 12:03, 26 February 2020 (UTC)[reply]
- @Jo-Jo Eumerus, if you mean the "From lack of discussion, this nomination appears to have no quorum" part, the intent was to give some context for why the task posts to some AfDs but not others. Otherwise open to tweaks on the default language. czar 02:20, 27 February 2020 (UTC)[reply]
- Yeah, I was meaning that part. Perhaps what I am thinking is that a bot that generally reports the existence of past AFD, PRODs, undeletions etc. to an AFD is more useful than one that merely informs you of the softdelete eligibility, but they are not exactly the same bot task. Jo-Jo Eumerus (talk) 07:12, 27 February 2020 (UTC)[reply]
- Surely, I can provide both them as the report shown. Maybe we can do some trial edits and see which is the most acceptable? --Kanashimi (talk) 05:16, 28 February 2020 (UTC)[reply]
- @Kanashimi: How long of a trial would you prefer? --TheSandDoctor Talk 21:02, 7 March 2020 (UTC)[reply]
- Thank you for reply. One week is good, and I will improve the code if needed. --Kanashimi (talk) 21:29, 7 March 2020 (UTC)[reply]
- @Kanashimi: How long of a trial would you prefer? --TheSandDoctor Talk 21:02, 7 March 2020 (UTC)[reply]
- Surely, I can provide both them as the report shown. Maybe we can do some trial edits and see which is the most acceptable? --Kanashimi (talk) 05:16, 28 February 2020 (UTC)[reply]
- Yeah, I was meaning that part. Perhaps what I am thinking is that a bot that generally reports the existence of past AFD, PRODs, undeletions etc. to an AFD is more useful than one that merely informs you of the softdelete eligibility, but they are not exactly the same bot task. Jo-Jo Eumerus (talk) 07:12, 27 February 2020 (UTC)[reply]
- @Jo-Jo Eumerus, if you mean the "From lack of discussion, this nomination appears to have no quorum" part, the intent was to give some context for why the task posts to some AfDs but not others. Otherwise open to tweaks on the default language. czar 02:20, 27 February 2020 (UTC)[reply]
- Approved for trial (7 days). Please provide a link to the relevant contributions and/or diffs when the trial is complete. I think this is one of those cases where a trial is needed both to demonstrate proof-of-concept as well as garner input from those directly impacted by the edits as to the usefulness of this task. Primefac (talk) 19:53, 15 March 2020 (UTC)[reply]
- An AFD I started, Wikipedia:Articles for deletion/Hazel College, Indiana, appears to the be first one this is tested on, and I strongly support it, due to abuse of prod removal I've seen and admins' occasional unwillingness to delete an expired prod. This is an excellent idea, provided admins are actually willing to regularly close no quorum discussions as delete. This one did actually get a delete vote after the bot's post but it was relisted soon after. Reywas92Talk 18:38, 21 March 2020 (UTC)[reply]
- @Czar: It seems that this is not a coding problem. Are there any good way to solve this? --Kanashimi (talk) 20:24, 21 March 2020 (UTC)[reply]
- @Kanashimi, I view this as a user error, not something to fix in the bot. The editor relisted the discussion when it was eligible for a soft deletion. If I had caught it, I would have notified the relister and closed it as soft delete nonetheless. The relister has a few messages on their talk page about incorrect closures. czar 23:34, 26 March 2020 (UTC)[reply]
- Thank you --Kanashimi (talk) 00:15, 27 March 2020 (UTC)[reply]
- @Kanashimi, I view this as a user error, not something to fix in the bot. The editor relisted the discussion when it was eligible for a soft deletion. If I had caught it, I would have notified the relister and closed it as soft delete nonetheless. The relister has a few messages on their talk page about incorrect closures. czar 23:34, 26 March 2020 (UTC)[reply]
- @Czar: It seems that this is not a coding problem. Are there any good way to solve this? --Kanashimi (talk) 20:24, 21 March 2020 (UTC)[reply]
- The AfD Wikipedia:Articles for deletion/Effort Administrator seems to have been one of those picked up in the trial. The text of the message Cewbot left is not clear. Firstly, "It seems no previous PRODs, previous..." I think should actually read, "It seems that there have been no previous PRODs, previous..." Secondly, the "It seems" language is odd. If I as a human editor were to post that type of denial it is a natural hedge against my capacity for error. I would expect a bot, however, to not need to make such qualifications. I would expect a bot to be coded in such a way that its search algorithms for those factors would be of a high confidence. That is, if a bot searches for something and doesn't find it, then there should be a high degree of certainly that the missing thing is actually missing. Saying "It seems" in bot results suggests the bot's search shouldn't be relied upon by closers and if that is the case, why should the closer be reading the bot's notice at all? Eggishorn (talk) (contrib) 20:29, 25 March 2020 (UTC)[reply]
- Thank you for reporting. I change the words to "There are no previous..." Since there are coding modified, I will trial edit more days. --Kanashimi (talk) 22:26, 26 March 2020 (UTC)[reply]
- "It seems" was a choice based on the bot not being certain that indeed there were no prior PRODs, right? Is the detection currently sufficient to say that it handles that case now (searching the diffs and edit summaries for prior PRODs)? czar 23:34, 26 March 2020 (UTC)[reply]
- @Czar: How about change to "After searching logs and edit summaries, it seems that there are no previous PRODs. And there are no previous AfD discussions, previous undeletions, ..."? --Kanashimi (talk) 00:15, 27 March 2020 (UTC)[reply]
- @Kanashimi:, I think you're kind of missing an important point in trying to adjust the verbiage: Is the code sufficiently accurate the bot to say: "I have not found X so X doesn't exist."? If the bot's results have to be qualified with anything that approximates: "I think X doesn't exist," then the code may not be accurate enough to be trusted. If they can't be trusted, the is no reason to have the bot do this task at all. That is, if an admin can see the notice and has to say to themselves: "This bot thinks that X doesn't exist but I have to check myself," then the bot isn't saving any work or replacing any effort. Its runtime is literally wasted. Eggishorn (talk) (contrib) 00:40, 27 March 2020 (UTC)[reply]
- @Eggishorn: The bot will search logs, edit summaries, AfD discussions, undeletions and redirects. The AfD discussions, undeletions and redirects are deterministic, but logs and edit summaries are not. Perhaps we have a better way to explain the meaning. Do you have any good idea? --Kanashimi (talk) 02:35, 27 March 2020 (UTC)[reply]
- @Kanashimi:, I understand where the possible error sources are now that you explained that. Is there any reason to think that I as a human can do a better job of searching the logs and edit summaries than your bot? That will change how the bot should notify AfD closers. If the answer to that question is "yes" then the bot should say something like: "This nomination has had limited participation and falls within the standards set for [[WP:NOQUORUM|lack of quorum]]. There are no previous PRODs, AfD discussions, current redirects and no previous undeletions have been located. This nomination may be eligible for soft deletion at the end of its seven-day listing." If the answer to that question is "no", then it should say: "This nomination has had limited participation and falls within the standards set for [[WP:NOQUORUM|lack of quorum]]. There are no previous PRODs, AfD discussions, previous undeletions, or current redirects. This nomination may be eligible for soft deletion at the end of its seven-day listing." I hope that helps. Eggishorn (talk) (contrib) 05:51, 27 March 2020 (UTC)[reply]
- @Eggishorn: The bot will search logs, edit summaries, AfD discussions, undeletions and redirects. The AfD discussions, undeletions and redirects are deterministic, but logs and edit summaries are not. Perhaps we have a better way to explain the meaning. Do you have any good idea? --Kanashimi (talk) 02:35, 27 March 2020 (UTC)[reply]
- @Kanashimi:, I think you're kind of missing an important point in trying to adjust the verbiage: Is the code sufficiently accurate the bot to say: "I have not found X so X doesn't exist."? If the bot's results have to be qualified with anything that approximates: "I think X doesn't exist," then the code may not be accurate enough to be trusted. If they can't be trusted, the is no reason to have the bot do this task at all. That is, if an admin can see the notice and has to say to themselves: "This bot thinks that X doesn't exist but I have to check myself," then the bot isn't saving any work or replacing any effort. Its runtime is literally wasted. Eggishorn (talk) (contrib) 00:40, 27 March 2020 (UTC)[reply]
- @Czar: How about change to "After searching logs and edit summaries, it seems that there are no previous PRODs. And there are no previous AfD discussions, previous undeletions, ..."? --Kanashimi (talk) 00:15, 27 March 2020 (UTC)[reply]
- "It seems" was a choice based on the bot not being certain that indeed there were no prior PRODs, right? Is the detection currently sufficient to say that it handles that case now (searching the diffs and edit summaries for prior PRODs)? czar 23:34, 26 March 2020 (UTC)[reply]
- Thank you for reporting. I change the words to "There are no previous..." Since there are coding modified, I will trial edit more days. --Kanashimi (talk) 22:26, 26 March 2020 (UTC)[reply]
- Thank you Fixed --Kanashimi (talk) 20:44, 27 March 2020 (UTC)[reply]
{{BAG assistance needed}} Since there are some concerns, May I trial edit more 7 days? --Kanashimi (talk) 00:20, 27 March 2020 (UTC)[reply]
- Approved for extended trial (14 days). Please provide a link to the relevant contributions and/or diffs when the trial is complete. I went through a couple hundred WP-space edits by the bot and didn't see any of the AFD edits, but I suspect this is largely due to edits made for other tasks. In the bot's edit summaries please link to this BRFA. When the trial is complete please list (or link to a list) of the bot's AFD edits, and use {{BotTrialComplete}} to indicate the trial has finished. Primefac (talk) 18:05, 29 March 2020 (UTC)[reply]
- Thank you I will use a tag. --Kanashimi (talk) 21:30, 29 March 2020 (UTC)[reply]
- Trial complete. It seems no big error, and the mistakes were all fixed. --Kanashimi (talk) 03:24, 12 April 2020 (UTC)[reply]
- @Kanashimi, the trial's over and I already miss the bot. :) I have a few thoughts. (1) Question: Does the bot detect former PRODs with high confidence? I didn't check for that before and wanted to confirm whether closers can rely on the bot for thoroughly checking the edit history. It's the difference between the bot saying that the nom "appears eligible" and "is eligible". (2) Is the bot's text (as discussed above) editable on-wiki, or is it hard-coded? (3) I still only see value in dumping the creation/deletion logs into the AfD text when explaining why the article is NOT eligible for soft deletion, because otherwise it's usually just a creation link, which adds little. I also still think there is value in possibly pursuing this type of abbreviated log posting for all AfDs but that's another discussion/different scope. (4) Can the edit summary be shorter? "Informing the article's PROD eligibility: Seems eligible for PROD" >> "assessed as eligible for soft deletion" vs. "assessed as NOT eligible for soft deletion" czar 04:28, 18 April 2020 (UTC)[reply]
- @Czar:
- As mentioned above, the logs and edit summaries are not deterministic.
- I may let it editable on a wiki configuration page.
- Yes, it is another discussion.
- Surely. I may let the edit summary editable on a wiki configuration page, too. --Kanashimi (talk) 05:28, 18 April 2020 (UTC)[reply]
Approved. Minor changes to edit summaries or inner workings do not need any approval, though any major changes should get approval via the talk page. Primefac (talk) 02:01, 22 April 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 WT:BRFA.