Topic on Talk:Article feedback

Jump to: navigation, search
Ziko (talkcontribs)

Hello, I just saw that the article en:Justin Bieber is rated only 3.24 with regard to completeness. I did not read the article thoroughly, but I am surprised that such a long (and most certainly very complete) article has no better rating. It seems that some readers' attitude to the article's object have caused that. This shows a not so mature behaviour of some readers.

Wasbeer (talkcontribs)

Yep, that is because people will base their ratings on the subject of the article and not on the article itself. If you give it enough time and data it will become a rather accurate popularity rankingsystem for BLP subjects, not an accurate rankingsystem for articles.

Bensin (talkcontribs)

I'd be very interested to hear a response to this from the proponents of the project.

DarTar (talkcontribs)

Yes, that's an interesting point partly related to my previous comments on trending topics vs. quality. We definitely need to think of how to cope with these effects. I am persuaded the solution should not apply to the rating collection itself but to how we calculate aggregate scores: for example we should filter out suspicious patterns that indicate that an article is being rated as a function of the topic's popularity in the media. Having more data on multiple ratings by the same users or IP addresses should also allow us to study the reliability of ratings. Unfortunately, because we are still using a small sample of articles, we have very little data on people who rated multiple articles.

Wasbeer (talkcontribs)

The easiest solution is to not ask for ratings but instead asking to comment on the article. Those comments can be displayed on the talkpage.

If, for example, someone gives the Bieber article 5 stars in every category, it means "OMG Justin Bieber is so cute I love him!". If that person wrote that in English on the talkpage then we would be able to understand the reason why that person rated that article in that way.

The same principle applies to 1 star ratings in every category, they basically mean "Fuck Bieber".

We should not ask for ratings, we should ask why people rate an article the way they do.

We shouldn't invent a problem (not being able to determine the intentions of the person who gave the feedback), then try adjust algorithms so the effect is less extreme.

Doug Weller (talkcontribs)

Wasbeer, if someone wrote that on the article talk page it should be removed. Talk pages aren't forums and I hope that this doesn't encourage people to use them as forums. Dougweller 07:45, 9 July 2011 (UTC)

This post was posted by Doug Weller, but signed as Dougweller.

Ziko (talkcontribs)

Any solution of the feedback problem that involves the Talk Page is *not* a solution. Usually people don't know about that page, and it is very difficult to edit it. (Yes, it is!) I like the Polish Wikipedia comment function. But this has little to do with this feedback function. Maybe it would be good to explain to people how to rate; that some typos don't mean that the article cannot have five stars.

Timl2k4 (talkcontribs)
Any solution of the feedback problem that involves the Talk Page is *not* a solution. Usually people don't know about that page, and it is very difficult to edit it.

Wrong, wrong, wrong! So wrong in fact I said it three times. The talk page is EXACTLY the place for constructive article feedback. If people don't know about it, that needs to be fixed. It it is too difficult to edit, that should also be fixed.

I bolded constructive, because any other feedback is useless.

Wasbeer (talkcontribs)

If you have a "feedbacktool" that allows people to leave a comment in a free text field on the article you can display the feedback on the talkpage.

Timl2k4 (talkcontribs)

That's pretty close to what I was thinking, the question is how to display the feedback and filter out garbage. Perhaps more seasoned Wikipedians could pick out specific feedback items and move them to the main discussion. Or at the very least take the feedback into account when trying to improve the article, without having to work off of vague ratings like the AFT would produce.

Ziko (talkcontribs)

Yes, that's what the Polish have. And it works; the comments are easy to fill in and are sent to a wiki page where experienced Wikipedians look at it. But our usual talk pages - they don't work. :-)

Bensin (talkcontribs)

I like this too. Asking for constructive criticism of the article is immensely better than mindless rating.

He7d3r (talkcontribs)

Yeh, I like it too! As mentioned here that feature (pl:MediaWiki:Wikibugs.js) is also in use on Spanish and Portuguese Wikipedias, and was already proposed on English Wikipedia. See:

Wasbeer (talkcontribs)

OK, is there anyone opposed to having a "feedbacktool" that allows people to leave a comment in a free text field on the article and displaying the feedback on the talkpage?

I would love to get more opinions on this, negative or positive. I honestly think this may be a good solution, and an improvement over the current situation.

Jorm (WMF) (talkcontribs)

This is the direction I have wanted this feature to go since its inception. However, the underlying technology required to make this feasible remains to be built. As such, we have not included it in the roadmap.

There may be intermediate steps that can be taken, but there hasn't been a great deal of hard thinking about it. A primary concern is one of spammy-ness: if anyone can leave a suggestion at any time, we'll start getting crap data almost instantly.

One way to gate this information flood is to ask for suggestions after the user has rated the article, and within a set context.

For example, if a reader rates an article's completeness low - say, a value of 2 - then we can ask them: "We see you have given this article a poor rating for completeness. What do you think could be done to make the article more complete?" and then give them a list of suggestions - checkboxes, say:

  • The article needs more photographs
  • The article needs more references
  • The article needs to be brought up to date with current events
  • etc.

And they could then enter a text comment after, if need be. This data would then be tallied and thrown into a "to do" list for the article, that could be actionable by editors.

The overall assumption that follows "if people want to make suggestions, they should use the talk page" is an untenable one. Readers simply will not do this because the editing interface in general and talk pages in specific is frankly terrible verging on the unholy.

One of the largest problems facing the continued existence of Wikipedia (and its sister projects) is the simply terrifying learning curve that new users are required to climb for even the most minor of interactions. Reducing this complexity even a little bit will help to offset the harshness when and if the user ultimately decides to engage in a more meaningful way.

Wasbeer (talkcontribs)

Among us are very smart people with great programming skills. You mentioned Catrope, I was referring to Krinkle earlier.

I understand that it is easier to include stuff in the roadmap when the underlying technology has been built already, but because we all agree this is the best way forward it may be a good idea to ask those people if they are willing to make this possible.

Of course they are volunteers (even if they get paid to do stuff like this they still have the option to decline the request), but we may be able to convince them to build the underlying technology in a short timespan if we offer them huge amounts of stroopwafels.

It would be unfair to ask them to do it in just 2 days, but if we postpone the full deployment a little bit and have a quick brainstorming session about how and when to ask for text comments I think it is possible to achieve this goal in an acceptable timeframe. Once the underlying technology is built we can make changes to how and when to ask for textcomments and compare the results.

I think this is a very important step and I am quite sure this will greatly improve the AFT. How do you feel about it? Is this acceptable to all parties?

Jorm (WMF) (talkcontribs)

The kind of technology we are talking about will take much longer than a couple of days to develop, and will be primarily back-end in nature.

This is not something that is worth postponing the deployment for.

Wasbeer (talkcontribs)

Well, I think you mean something that is much more complicated than what I have in mind.

I think you are thinking about something with a backend like File:Article feedback extended review - Review list.svg. That would be a lot of work of course.

I was thinking about just introducing a simple textbox where people could leave a comment, press a button, and the comment would be added to the bottom of the talkpage like so:

==Feedback by ~~~~==

This article is basically OK but I'm afraid source X is given undue weight. ~~~~

What is your opinion on that?

He7d3r (talkcontribs)

Another approach to getting data about a page is proposed on w:en:Wikipedia:Sticky notes. See the image on the right.

Reply to "Justin Bieber article not complete?"