Jump to content

Community Wishlist Survey 2017/Reading/Pop-up showing authorship info

From Meta, a Wikimedia project coordination wiki

Pop-up showing authorship info

  • Problem: It is often difficult to assess Wikipedia articles. Some are the result of multiple authors collaborating, others are the efforts of one PR agent, an editor and an Arb joining forces to validate an essay on a CEO's activism, still others are the Herculean efforts of single accounts. Readers should be able to find this information quickly. Responsible scholarship gains authority (at least in the humanities) due to the reputation of its authors, and yet the history pages on Wikipedia do not give a global overview of the "table of authors" who wrote the article whose history they record typo by typo. Recently, questions have been raised about the WMF's compliance with a court decision requiring that covert advertising be clearly identified as such.
  • Who would benefit: Critical readers, the WMF
  • Proposed solution:
  1. Add a pop-up version of the "Article Info" page at xtools to all visible pages on the project (or at least to all articles in mainspace). This tool is several clicks away from the reader at the moment. (view history > revision history statistics)
  2. Improve the "Article Info" algorithm so that it discounts reverted text and reversions from the edit totals. (Cf. the graph for Donald Trump or Hillary Clinton to see more clearly why the resulting pie charts are skewed... basically, the top editors are those who revert page blankings).
  3. develop color codes for degrees of collaboration that appear in the article itself.

This would probably not be enough to address the serious concerns raised about compliance with European Fair Trade Laws in OLG München · Urteil vom 10. Mai 2012 · Az. 29 U 515/12, but I believe it's a small step in the right direction.

  • More comments: I just learned that User:Doc James tried to get this done years ago and failed due to technical problems and a lack of coders able to make it work. How about we make it happen this time?
  • Phabricator tickets:

Discussion

[edit]
Yes, two requests with over 60 supports each, it does look like German Wikipedia is leading the way on the question. Thanks for the info. SashiRolls (talk) 21:30, 15 November 2017 (UTC)[reply]
  • Re: "This tool is several clicks away from the reader at the moment. (view history > revision history statistics)". It's not quite what you're asking for, but just thought I'd mention that the XTools gadget makes the articleinfo page available with one click (a link under every page title). And perhaps it wouldn't be too hard to add the author list directly there. Sam Wilson 06:58, 15 November 2017 (UTC)[reply]
Aha! Thanks for pointing this out. So the technology exists that allows any reader to do the first step of what I'm asking, and using the author line for statistics and a link to authorship seems like a reasonable first step. Why isn't this standard? That's so much better! Failing that, a link on the sidebar to a (just the facts no promo) "reader's guide to wikipedia" would be useful so that first time visitors could learn about how to find author info, COI info, how to install the author-info gadget (not obvious for someone who hasn't been here for a while), where the disclaimers are, why they're important, etc. SashiRolls (talk) 21:18, 15 November 2017 (UTC)[reply]
Upon reflection, you can only install/activate a gadget if you create an account, so really none of this has any effect on the general reader who does not have an account (and the non-member reader ("client" of "knowledge as a service") is the target for the suggested improvement). SashiRolls (talk) 00:16, 21 November 2017 (UTC)[reply]
  • Improve the "Article Info" algorithm so that it discounts reverted text and reversions from the edit totals – It is doing this to some extent, but we're working to improve the logic, see phab:T179996. As Sam said above, there is a gadget that will give you some high-level information at the top of each page, with a link to the full statistics. This makes it one click away instead of two. Is that sufficient?
    With T176912 we looked into reviving the WikiHistory gadget, which would give you "text shares" (authorship percentages) in real-time. However getting this information requires parsing every single revision within the article. It is not feasible to do this on every page you view, in real time. WikiHistory was able to do this because there was a bot that precomputed the data. In addition to significant maintenance burden, the downside to the bot was it had to be enabled for each wiki individually, and that was assuming that community would actually make use of it (it wouldn't make sense to run a bot no one is using). It requires considerable resources to precompute and maintain this data. In my opinion this is not a good system. We can get you improved data that will help with the issues you are facing, but it should remain an on-demand service, and not an automatic service.
    The other thought is to build these stats directly into MediaWiki. For that I suppose we'd keep running totals as each revision is made, that way we can serve it to you very quickly. This would be a huge effort, and perhaps not worth the while given the size of the audience it would benefit versus development time. It would also most likely be bound by the upcoming revision refactor. In other words, I don't think implementing something like this in MediaWiki could happen in the short term.
    I have never heard of the European ruling you speak of, but I will let the legal team know about it. MusikAnimal (WMF) (talk) 16:36, 15 November 2017 (UTC)[reply]
Yes I've thought about the problem and realize that it's pretty tricky. I was assuming the bot to establish authorial "share" (color coded collaboration as of ##date##) would run on the periodic dumps not on the live database. I recognize too that it's not very environmental of me to ask for calculated author info to be made readable & accurate. Just to be clear: RickinBaltimore really is not the primary author of the Donald Trump article. He just reverted a page blanking. Glad to see you've already caught this and are working on it. Thanks!
I notice volunteers get BY credit for text they copy from paid editors who cannot edit pages directly. The average reader will still not find accurate authorial info even with the gadget unless they also look at the talk page (on en-wiki). Another example: Kosmos Energy (talk page) Unlike with Krzanich, full page protection wasn't deemed necessary to keep mad vandals from messing up / reverting the PR agency's prose. Making that gadget standard would be a good thing to push for. I realize this problem of misattributed edits is due to the en.wiki policy of embedding the COI template among all the other templates on the talk page rather than being on the article page. When policy gives you bad data, there's not much the developers can do. Thanks very much for your response. You've made some great tools. SashiRolls (talk) 21:14, 15 November 2017 (UTC)[reply]

There is an old research project called WikiCredit on this topic. As you can see there, meaningfully quantifying authorship is not an easy problem. --Tgr (WMF) (talk) 01:53, 20 November 2017 (UTC)[reply]

Okay we know have something working on EN WP using a script thanks to user User:Wurgl.


Example

The database is currently being build over the next few weeks. So for pages with lots of edits it can take up to 30 min to generate results. For short pages it works in seconds.

Copy and paste the following to here
mw.loader.load('//en.wikipedia.org/w/index.php?title=User:Wurgl/WikiHistory.js&action=raw&ctype=text/javascript'); Doc James (talk · contribs · email) 01:21, 27 November 2017 (UTC)[reply]

  • If the project is feasible, maybe a color aid could be implemented to show which text was added by which editors, such as Google Docs currently does, although I don't know the implications of this or how difficult it could be to apply in real time. --Jamez42 (talk) 13:06, 28 November 2017 (UTC)[reply]
  • To respond to TheDJ's oppose: while I understand that it may well be a technical challenge to generate accurate information, I do not believe that this suggestion caters first and foremost either to Wikipedians or to Sceptics, but to serious consumers of a collaborative reference work. It is entirely possible that it will become necessary in order for the projects to be legally compliant in some jurisdictions in the future, and would certainly be a significant help to anyone wishing both to assess trustworthiness & (possibly) to read other well-written articles by specific contributors. The point of suggesting a scroll-over pop-up in 1 & 2 was that it doesn't add anything more than the word "authors" to the screen. The idea of having "authors" on the lefthand sidebar is also more discrete than the line at the top of the page which has currently been proposed. The color coding I suggested for degrees of collaboration in 3 could be the background for the <span> containing the words "authors". Much less cluttery than an infobox for example (not that I have anything against infoboxes) SashiRolls (talk) 00:26, 30 November 2017 (UTC)[reply]
  • Comment Comment I share misgivings of User:SMcCandlish that this might lead to some non-healthy competition and WP:OWN-oriented approach to articles. I also do not like added clutter. On the other hand, I like the idea of some indicator of article trustworthiness and I trust more articles with a lot of authors, so seeing hat 90% of an article come from a single contributor, could be a warning about lack of vetting. However I would prefer that as an opt-in gadget. --Jarekt (talk) 13:49, 4 December 2017 (UTC)[reply]
    • Yes, this is a relevant concern. The idea behind the color-coding was to mark primarily single voice articles as such (as potentially needing to be read with a grain of salt). This just makes visible outside what is already visible inside (COI editors may well already send their potential clients pages of links to the xtool profiles of their greatest hits). SashiRolls (talk) 02:35, 5 December 2017 (UTC)[reply]
  • Very neat that the WikiHistory gadget was revived! I doubt this proposal will make it to the top 10, but just so it's clear, I don't think WMF will build any kind of gadget like the WikiHistory one. We can however improve XTools authorship detection, and/or build another tool for this, as explained at Special:Diff/17426787. I'm sorry we didn't interject and ask the proposal be reworded, because I suspect people would have less concerns about WP:OWN if the authorship statistics were one at least click away, and not shown automatically atop every page. MusikAnimal (WMF) (talk) 18:10, 8 December 2017 (UTC)[reply]

Voting

[edit]