Wikidata talk:Mismatched reference notification input

In order to understand your needs and workflows better, we would love to get your feedback on this page. Feel free to create a new section to answer the following questions. You can also leave comments in the “other discussions” section.

Questions

• (1) While testing the feature, did you notice problems or things that are hard to understand?
• (2) In what part of your daily work on Wikidata would you imagine this feature being useful?

While creating that feature we thought of making it persistent, which means that the notification would stay on the Item page, and other editors would be able to see it, as well as you would see the notifications on statements where other people edited a value without changing the reference.

• (3) Do you think this would be useful in your workflows and if yes, how?
• (4) Do you see any problems that could occur?
• (5) Which additional features would you need in order to improve the matching between values and references on Wikidata?

Salgo60

• (1) No problem so far
• (2) its easy missing things when updating data manually so this is excellent
• (3) yes I use the gadget DuplicateReferences a lot so question is how they work together
• (4) not now
• (5) I recommend listen to Denny Vrandečić link he has a lot of visions

- Salgo60 (talk) 13:16, 8 January 2020 (UTC)

Ainali

• (1) No.
• (2) Not making involuntary mistakes.
• (3) Yes, for catching mistakes.
• (4) That I miss the notification and leave the item without correcting a mistake I made.
• (5) I would suggest giving a popup dialogue before publishing: "You changed a value, but the reference remained the same. Do you want to publish?" to help me from not doing the mistake in the first place.

ArthurPSmith

• (1) No problem. However, I was expecting to see the popup notification immediately, rather than the (!) sign of an issue that you then have to click to see the problem. I think it would be better to make the notification more in-your-face, requiring an additional click to dismiss it. (just at the point in time of the edit, it can hide like a regular warning later if it is persistent at all).
• (2) One doesn't always notice that there is a reference on a value, so this is a useful notification to check that your change conforms with the existing reference, or if not to change the reference or add a new one - or not make any change if your proposed change is in error. A common example for me is employer statements on items for scientists - these often are referenced to ORCID, so it is good to be slightly nudged to confirm that whatever you are changing conforms to that reference - or to look into the issue further if not!
• (3) I assume this is asking about persistence of the warning. It would be nice if it could be dismissed only by autoconfirmed editors, so that changes made with no reference update by IP or new users would stand out more clearly; I think that would be helpful in patrolling. In general having a persistent warning is useful here to indicate somebody should confirm the value.
• (4) It seems like the warning is not there if there is ANY change to the references - for example adding a new reference, or modifying just one of several existing references. The warning is also not there if you add an additional value for a property that already has a referenced value. This is all probably fine, but it may make it a little less useful for patrolling if vandals find out how easy it is to avoid generating the warning...
• (5) The popup dialogue suggested above before publishing might be helpful; alternatively maybe what I suggested above, a popup warning after publishing. But this would be fine to use as it is now. If it is to be persistent then there definitely needs to be a way to dismiss the warning ("I've checked the reference and confirmed the new value" or something...)

Jheald

• (5) A link to the diff might be useful, if/when the warning is made persistent (which I think would be valuable, at the 'suggestion' level of warning). Jheald (talk) 11:42, 13 January 2020 (UTC)
  • Although, with due respect to User:Moebeus below, the team probably are correct to be introducing this in baby steps first, at least initially. Subtle (ie not overtly obvious) modification/vandalisation of referenced content is indeed something that it would be good to move on, as soon as we can. But there are also legitimate changes that can create false positives -- eg if the object item of a statement is merged, or an approximate date is re-coded, or additional information is added from the source by way of qualifiers. So trying to get a more comprehensive understanding of these before deeper roll-out may make some sense. Jheald (talk) 09:08, 17 January 2020 (UTC)

Moebeus

• (1) Nope, dead easy.
• (2) Everyday, all the time.
• (3) 100% useful. As a best-practise-reminder to always check/update references.
• (4) Nothing that comes to mind.
• (5) This is a great step in the right direction, very nicely done. My only comments are directly parroting Jheald : It would be even better if made persistent with a link to diff, with an option to dismiss tied to some kind a "patrol-action". Moebeus (talk) 01:03, 17 January 2020 (UTC)

Jarekt

• (4) One issue I run into is when I first updated reference and then the value here triggering this flag. --Jarekt (talk) 00:10, 4 February 2020 (UTC)

Edoderoo

• Many properties have a value changing regularly, while the source/reference stays the same. Then this "warning" makes no sense. Examples are: weekly tennis balance single/double (P564/P555). This warning should be optional, and not default for all properties. Edoderoo (talk) 20:21, 4 February 2020 (UTC)

ℇsquilo

• (1) No.
• (2) More irritating than useful. Full reference is visible when changing the value anyway. Often the reference is correct, but the value is misinterpreted.
• (3) No.
• (4) Pass.
• (5) Pass.

Ah, today it has the added feature that it can be removed with a click. Much appreciated. /ℇsquilo 07:46, 6 February 2020 (UTC)

Finn Årup Nielsen (fnielsen) (talk)

1. No
2. All over
3. Yes
4. Accidentally pressing remove
5. It might be interesting to have it extended to other situations: "There are no reference, but this statement should really have one" (citation needed), "the source is probably ok, but a better source would be nice" (e.g., paywalled), "the source is wrong", "the source does not support the statement".

— Finn Årup Nielsen (fnielsen) (talk) 11:02, 9 April 2020 (UTC)

Other discussions

• Since when does test.wikidata have an "Automatic" option for adding a reference? This is really nice! ArthurPSmith (talk) 19:30, 8 January 2020 (UTC)
  • Hehe that is the integration of Citoid. It's using the same backend that supports easier referencing on Wikipedia. It's unfortunately not quite ready yet for handling Wikidata. We're working on it with Marielle from the WMF. --Lydia Pintscher (WMDE) (talk) 20:10, 8 January 2020 (UTC)
• I would find it useful to get rid of false positives after changing the value. E.g. fixing the source and WD together (imported from WP stays correct), replacing e.g. located in the administrative territorial entity (P131) based on coordinates by a more precise information, etc. --Herzi Pinki (talk) 17:30, 3 February 2020 (UTC)

Hello @Herzi Pinki:, and thanks for your feedback! I have a few questions, to make sure that I understood correctly.

In the first case, are you talking about using a Wikipedia page as a source, then you would update the Wikipedia page and the value? We assumed that in most cases, per community rules, Wikipedia should not used as a source. But maybe this case is more frequent than what we thought.

If you'd update the coordinates based on a more precise information, wouldn't you update the reference in the same time?

Thanks, Lea Lacroix (WMDE) (talk) 08:24, 4 February 2020 (UTC)

There is one thing behind. I do not trust in references at Wikidata, imported from any Wikipedia is problematic, more, if coordinates in Austria are imported from the mongolian WP. Coordinates taken from GeoNames is a bunch of unpreciseness. Elevation information (above sea level) derived from a geomodel using wrong coordinates makes differences up to 1000 meters in height. Yes, Wikipedia imho is not a reliable source, OSM is neither a reliable source, Geonames is not a reliable source. Mass imports from almost any collection of data will also import erroneous data. So why should I change a single reference when million references are rubbish? --Herzi Pinki (talk) 11:21, 4 February 2020 (UTC)

• At the moment I can remove the reference and the warning will go away. Maybe we should enforce references for all properties. If the value is changed and the reference is deleted, the value should only be saved after adding a new reference. --Herzi Pinki (talk) 17:33, 3 February 2020 (UTC)
• ...

• Suggestion: note it in edit history. --- Jura 10:18, 6 February 2020 (UTC)

• I was expecting to see the warning when I made this edit, but it did NOT appear. That edit was made after I had done several other edits on the page without reloading it - a javascript bug of some sort? ArthurPSmith (talk) 13:23, 6 February 2020 (UTC)

• I think it should look different from constraint violations --- Jura 13:27, 6 February 2020 (UTC)

• On Openwall Project (Q4045922) I just updated http to https keeping the "imported from enwiki" as is, it worked as designed, i.e., I ignored the warning, because I actually exported the updated URL from enwiki to WD, and re-imported it on enwiki using {{official website}}. –84.46.52.123 15:52, 9 February 2020 (UTC)

Wrap-up (5.3.2020)

Hello all,

Thanks a lot for taking the time to test the mismatched references feature and giving feedback! Over the past two months, we collected and analyzed your comments in order to understand your needs and move forward with the development of the feature.

Here’s what we learned:

• The feature generally received positive appreciation, accompanied with suggestions for improvement
• Several people mentioned the persistence of the feature as a useful thing to have. We will focus on enabling this possibility for the next steps of the feature.
• Some of you mentioned the idea of warning the editor before publishing a change without updating the reference. The suggestions provide different options, from “display information” to “prevent the edit”. We will certainly look in that direction, although we want to be careful of not adding too many layers of confirmation buttons or warning pop-ups in the editing process.
• You described a lot of different workflows and false positive cases, which made us think about how we can highlight a potential problem (mismatched reference) without getting in the way or slowing down the editing process.

This feedback loop will stay open, you can continue giving feedback on this page whenever you want. The development team will keep working on the feature later in 2020, I will announce the next steps through the usual communication channels.

Thanks again for your feedback! Lea Lacroix (WMDE) (talk) 07:53, 5 March 2020 (UTC)

remove the reference

should be an one klick option. "German Wikipedia" was no valid reference from the beginning. --Bahnmoeller (talk) 09:41, 12 March 2020 (UTC)

adding qualifiers

I modified a referenced statement earlier today by adding a qualifier, and that didn't seem to trigger this popup warning. Not sure I'd tested that before, but it seems maybe it should...? ArthurPSmith (talk) 18:24, 9 April 2020 (UTC)

Update (27.4.20)

Hello all,

Thanks for the additional feedback, that we took in account for future developments of the feature. We will soon remove the link to this feedback loop from the pop-up, but this page will stay open for comments and suggestions.

We will keep working on the feature later this year, in order to integrate some of the suggestions that you provided. Cheers, Lea Lacroix (WMDE) (talk) 11:01, 27 April 2020 (UTC)