Wikipedia:Arbitration Committee/Requests for comment/Article creation at scale/Closing statement

Closing statement for Article Creation at Scale RFC. Panel close by:

• -- RoySmith (talk) 20:41, 4 November 2022 (UTC)[reply]
• Katie^talk 22:51, 4 November 2022 (UTC)[reply]
• TheSandDoctor ^Talk 20:53, 4 November 2022 (UTC)[reply]

Technical assistance from Firefangledfeathers is gratefully acknowledged.

Fails.

The way this question was worded, it wasn’t clear how the results should be scored, so we needed to improvise. There were 7 items (A, A2, B, B2, C, C2, and D) which were to be ranked in order. We gave 7 points for a first-place ranking, 6 for a 2nd place, etc down to 1 for a last place. If a choice wasn’t ranked at all, we gave it 0 points. Those of you who got creative in your rankings can consider yourself trouted; we did our best to map your prose into numerical scores.

The raw data is available on Google Drive (still trying to figure out how to get WikiMedia to upload an xls file).

A raw summation of the votes for each choice would show that “D” was the most popular, by a substantial margin. However, we’re unsure if that accurately reflects the consensus, per Vote splitting.

It might be reasonable to consider C and C2 as a group (and likewise with the A’s and B’s). It seems likely that somebody who voted for C2 would be happier with C than with D. If you look at things that way, then the answer is “The C group got the most votes, and within that, C was substantially more popular than C2”.

@Xeno and Valereee: We request that the moderators conduct a 7-day run-off, with the simple binary question, “Should we enact C?”, no commentary allowed.

As an aside, if a multiple-choice question like this is to be used in the future, we strongly recommend that a specific algorithm for scoring be established before the RfC begins. Leaving it to the vote counters to decide how to count the votes is a recipe for disaster. See Ranked voting for a discussion of some possibilities.

Enacting option C is the clear consensus of those who responded during the runoff. Looking at just the responses from people who participated in the original RFC, the numbers are closer, but it's still consensus to enact C.

• -- RoySmith (talk) 00:29, 13 November 2022 (UTC)[reply]
• TheSandDoctor ^Talk 04:12, 13 November 2022 (UTC)[reply]
• Katie^talk 01:02, 13 November 2022 (UTC)[reply]

This question got essentially no responses.

Fails almost unanimously.

Fails on numbers, and also the support votes are mostly wishy-washy “Something like this, but not exactly what was proposed”.

Passes by a wide margin. It’s clear people want some sort of definition, even if they can’t agree on what it should be.

Fails overwhelmingly.

Fails by a wide margin.

Fails by a wide margin. The proposal failed to quantify how "quality" would be measured.

Withdrawn.

Moved to the second RfC.

Moved to the second RfC.

Fails by a wide margin. Parties noted that it is pointless and consider deletion to not be a net benefit, especially considering that quarry queries can be made to get the same results.

Fails overwhelmingly, with reference to WP:NOTCOMPULSORY.

Moved to second RfC.

Fails by a wide margin.

Fails nearly unanimously.

There’s no question here.

Some of the major issues people talked about were

• Does it matter if it’s prose or not?
• What matters is whether it’s reliable, which may not correlate with whether it’s prose.
• Human-curated vs machine-gathered data?
• Databases vary widely in quality

We could not discern any consensus position, but closer to “fail” than “pass”.

On the surface, a majority for “pass”, but so few people responded here, we’re not sure we can legitimately divine anything.

This isn’t phrased as a question. We assume it was intended to say, “Should we develop…” In any case, while “support” got more votes, we find several of those to be unenthusiastic: “I wonder about the practicalities” “Difficult to implement, but I support in principle” “I do not think it will have any useful effect” “I don’t think it will solve this problem” We’re calling this a “pass”, but just barely, and noting minimal participation.

Technically passes on the numbers. The opposition has some valid points about BOTPOL being the wrong venue and the lack of a definition of how many article creations cause this proposal to take effect. A number of those opposed support in spirit but are opposed over the proposal's ill-defined nature.

We acknowledge that we were not adequately clear in our original closing of Q17 and are issuing the following clarification:

No adequately robust consensus was formed by this discussion that can be relied upon to change bot policy. While proponents debated what would trigger this proposal to take effect, those in opposition noted that human editing falls outside of the scope of bot policy^[a] and noted that the definition of “mass creation” was sufficiently imprecise/ill-defined so as to cause considerable problems with implementation.

Ultimately, this discussion was too close to adequately give a clear, actionable, consensus. As an aid for future discussion, here's a list of salient points we saw on the two sides:

Support:

• Early review would have prevented some ARBCOM cases.
• Even mass creations that are ostensibly human-powered are really semi-automated and don’t get much human oversight, so those processes should be reviewed.
• Talking about any mass-creation scheme ahead of time will save a lot of effort on the back end cleaning up problems.
• Eliminate the loophole of copy-pasting from off-wiki automated processes and then claiming it’s not automated.
• Any large-scale editing (whether automated or not) should have some kind of prior community review.
• Perhaps allow creating a small number of example articles (10?) and shop that around for approval before proceeding with a larger scale project.
• Some of the semi-automated schemes that mass-create articles ostensibly by hand would actually be done better if they were fully automated.

Oppose:

• Controlling human editing is out of scope for WP:BOTPOL
• No need to make this change, since BOTPOL already references MEATBOT
• Given that “mass creation” is ill-defined, it’s unworkable to enact policy which depends on that definition.
• How do you discern the dividing line between similar articles in a series that are truly human authored versus essentially automated processes that are just front-ended by humans who contribute little or no editorial input?
• This would be contrary to WP:BOLD.
• We should concentrate on quality not quantity; if somebody is cranking out many well-referenced articles a day, that’s a good thing.
• If somebody is creating low-quality articles, the solution is to block them.

Statement by:

• TheSandDoctor ^Talk 06:23, 11 November 2022 (UTC)[reply]
• -- RoySmith (talk) 16:16, 11 November 2022 (UTC)[reply]
• Katie^talk 13:25, 11 November 2022 (UTC)[reply]

Fails by a wide margin.