The Longest Now

To be honest, AFT always seemed to come across (particularly in the beginning) as being a project that people were trying to convince the community that they wanted, instead of the other way around. However I say that without really being a Wikipedian, or involved in AFT, so I don’t really know.

While there are certainly cases where tech projects can validly be of the form “convince users it is needed”, that is a dangerous road and one has to be careful when embarking on it.

I’m looking forward to your next thoughts on this.

I don’t know the answer to these questions. But it’s hard for this not to be one of my favorites so far: http://en.wikipedia.org/wiki/Special:ArticleFeedbackv5/The_Orion/885760

I do think we could have gotten a more realistic community assessment sooner in the process, with a much simpler version of AFT5. It’s hard to imagine that the outcome would have been different. (For the record, there’s still a trial underway in German Wikipedia, and one planned for French Wikipedia, both initiated by the communities of those projects.)

Fundamentally, a lot of users find the idea that there’s a new flood of low quality contributions they somehow have to deal with very discouraging. It might have helped to completely hide the feedback from logged out users, but I doubt it — it’s very hard to find as it is. Still, the idea that a nasty comment is left unmonitored somewhere in Wikipedia troubles the collective sense of hygiene.

I’m not sure on what time frame you are thinking about ‘outcome’ here.

I see people commenting that they think English Wikipedians “don’t like” or “won’t allow” AFT-like comments, but that doesn’t seem correct to me – those aren’t questions being considered by the community. Recentchanges — and the entire history of wiki contributions — involve various floods of contribution that somehow have to be dealt with [eventually].

Some long-term outcomes that seem reasonable to me:

At one level every page should have one-click feedback: annotation and commentary should be as easy as possible.

At another level, commentors should have instant feedback on their comment: they should be able to: see what it will look like in context…
edit it… follow further discussion of it… see the comments of others, cluster related feedback. They should have some sort of template. (“Add a new section” prompts for title and subject. AFT5 prompts with “e.g., This article needs a picture.”) All of these things help improve the quality of feedback, by example and expectation. Some of these things we want to do better for Talk pages generally; but it’s interesting to see that even simple things like “have an edit button” were dropped for the first AF5 version.

Similarly, everything should be easy to find in a single stream or feed of updates. That’s what wikis themselves are good for: unstructured and aggregated work. Talk pages copy that strength for discussion – and it allows them to be useful despite their many obvious weaknesses. Having a way to edit your input, and a way to comment on it, and a way to follow others’ comments, in the same place: another simple wiki trait. All of these things are /almost, but not quite/ possible with AF5 today.

At another level the spectrum of visibility shouldn’t go from “not enabled by software” to “visible to the public and permanently archived”. At least two layers of quarantine is sensible – content that any logged-in user can see [say, O(5M) live accounts], and that the submitter can see, but that others do not by default. And content that is hidden, and can only seen by reviewers [say, O(100,000) active editors]. Beyond this is deleted content that only admins can access [say, O(1,000) admins and stewards].

It is interesting that we have to continually revisit and argue these ideas. Some community members make the same arguments against making editing easy that wiki-detractors made ten years ago. And we have yet to figure out good balance/defaults for visibility issues, even for articles and images themselves.

Hey Erik and or all, did development on AFT2-4 (stars) switch pretty much to AFT5? I am curious because it seems like there are things that could make AFT2-4 more helpful (like seeing a little chart of feedback over time) that never seem to have gotten implemented.

I think longer term, this has the potential to weed out a lot of ‘spun’ articles out of say Google’s search results. Have you thought about say this application? Having users rate the quality of articles in Google’s search results?

I agree; it would be nice to see Google take that seriously. They tried this briefly, but it never got much attention or traction. MediaWiki could store a similar cache of searches and their most popular results, which would be a step in the right direction. -sj