Topic on Talk:Reading/Web/Projects/Related pages/Flow

There has been a lot of healthy conversation about this feature. Here is a summary of issues raised along with proposed responses:

Reading/Web/Projects/Related pages#Initial Community Feedback

Here is a proposal for moving forward:

Reading/Web/Projects/Related pages#Proposal for moving forward

These are just suggestions, so very curious to hear your thoughts.

Thanks to everyone who has given feedback or discussed so far, but in particular: @Gestumblindi @Alsee @SSneg @Ruud Koot @Matthiasb @Sadads @Jdlrobson @Finnusertop @Wittylama @Kusma

Have you considered posting a Request for Comments at Village Pump?

Hi Alsee- I have, but trust you and others on this list to tell me what the best path forward is.

If we did post an RFC, I would hope that we could co-write it to ensure we feel the framing is balanced. Let me know!

I greatly appreciate the very positive community engagement. Collaborating on an RFC is an idea I've floated before.

Per your comment above I've gone ahead and created Reading/Web/Projects/Related_pages/Draft_RFC. I'm not sure if it resembles what you had in mind. Don't hesitate to boldly change anything and everything, or even suggest something else from scratch. Everyone is invited to edit the draft or comment on its talk page.

Parts of the draft are essentially speaking with the Reading team's voice, but I copy-pasted as much as possible with limited modification. Aside from the general invitation aggressively edit the draft, please closely examine those bits and correct anything I got wrong.

Thanks, @Alsee! I appreciate you taking the first stab (I wouldn't know where to start), and will have a chance to review tomorrow.

Hey @Alsee, I just wanted to add that village pumps are not ignored. In fact. The first announcement about Related pages, was made in November, on the English Wikipedia Village pump. There were no responses to the post, and the conversation continued to happen in this more releavant page here, by several interested members from English and other Wikipedias. Hence, it made perfect sense to start the updates by directing the conversation to the already engaged community members.

@Alsee I took a stab at editing the RFC with some more context, but mostly feel it is solid and balanced. Thanks again!

Moving forward, I suggest considering adding a "Close (X)" button on the article cards to hide them if they are useless/inappropriate/unrelated, sort of like Google Banners and many other ads that are placed onto pages by algorithms that can be wrong. Most users won't notice it but the concerned editors will be able to use it to quickly block unrelated content. Then, potentially, this can be used to train the selection algorithms.

Interesting feature request! I like the notion of letting users vote with their actions, but this would require a lot of work - maybe a whole quarter and I don't know if we can justify it. Nonetheless, I just started a list on the project page of feature requests and added it there for tracking.

@Jkatz (WMF) I'm in the process of setting up a poll ("Meinungsbild") in German-language Wikipedia, see de:Wikipedia:Meinungsbilder/Mehr erfahren. According to German Wikipedia's rules, the poll needs 10 supporters before it can be held. Although I'm sceptical regarding the feature, I'm trying to gather a balanced set of pro and contra arguments - still, you'll notice that the page currently contains a lot more contra arguments (not all of them originally gathered by me). I really try to think of more points to add to the "pro" section to make it more balanced, but it's not that easy. Translating the current points:

Pro:

• The Related pages suggestions may encourage reading more about related topics.
• Associative links, if selected automatically, aren't really in contravention of Themenringe and Neutraler Standpunkt (NPOV), because there's no subjective selection by humans, but a dispassionate algorithm that simply suggests pages that seem similar.
• The feature is active in the Wikipedia app since January 2015, basically without protests.
• Users clicking on more pages are generating more page views, thus potentially more donations.

Contra:

• The "Related pages" links can be perceived as part of the article. For readers, it's not clear that these three links, unlike the actual article, aren't curated by humans, and an improper selection may be alienating.
• Manually correcting "Related pages" links may lead to disputes about a fitting selection, and a manual selection would be against Wikipedia:Themenringe.
• In cases where the links displayed are meaningful, they're usually already present at a fitting place in the article. Links that can't be integrated into the article are potentially rather not fitting.
• The embedded thumbnails don't allow for a direct link to the image description page, which may be problematic from a license point of view for images that aren't in the public domain (partly tracked in T124225 ).
• Through the algorithm choosing the "related pages" links, the Wikimedia Foundation interferes with content matters.
• The "related pages" links are a superfluous and imprecise duplication of the manually curated "See also" links (de:Hilfe:Siehe auch: Associative links are not suitable to create an unclear connection between an article's subject and connotated headwords such as "Far-right politics" or "Kitsch", bypassing the requirement to cite sources.)
• The "related pages" links visible in the article are invisible in the source code as well as in the Visual Editor; although it's possible to edit them, it's not obvious.
• The descriptions from Wikidata that are used for the short descriptions are disproportionately prone to vandalism which is then displayed on Wikipedia without being sighted [German Wikipedia has Flagged Revisions with "sighted versions"]

Neutral:

• We don't know whether the possibility to change the selection manually will lead to a new kind of vandalism and revert-wars, as some Wikipedians fear.

I think that "Users clicking on more pages are generating more page views, thus potentially more donations" portrays Related Pages as a "money-generating feature", a method of improving WMF's cashflow (which is skyrocketing anyway).

I think that "more clicks, more time on site" primarily means more value for readers, more curiosity satisfied, more Wiki-habit forming, more chance that the user finds something wrong or missing and fixes it, i.e. becomes an editor. And Wiki needs more fresh editors.

So I guess that any feature that keeps people here longer, is a good feature. And I also think that a 'pro' argument along these lines should be added to the discussion.

@SSneg Thanks - I added something along the lines of your suggestion to the German draft. "Any feature that keeps users longer on Wikipedia is a good feature. Besides the additional knowledge readers may gather, there is an increased chance that visitors will become Wikipedians, for example because they wish to fix errors they found."

@Gestumblindi Thanks for leading this and I appreciate you reaching out for comment. I agree that this is not about donations. @SSneg's interpretation and suggestion is a good one.

@Gestumblindi, possibly one more pro point, is supporting discoverability of orphaned articles.

@Melamrawy (WMF) Really orphaned pages (no links to the article whatsoever) seem to be rather rare in German-language Wikipedia. The special page "Verwaiste Seiten" (orphaned pages) at German Wikipedia lists currently only 128 pages, and some of them aren't actually orphaned anymore (as the list is based on yesterday's cache). Apparently, there are considerably more (and older) orphaned pages in English-language Wikipedia (at least several thousand, it seems; the first 500 hits of the special page end at Bucaramanga Nest, which is an orphaned article from 2013), but then, English Wikipedia is also considerably larger :-)

Still, there are of course lots of semi-orphaned articles with few links to them; so I think it's nevertheless a pro point that can be added. I'll add something along these lines: "Wikipedia users may discover articles that are orphaned or have only a few links leading to them."

An orphan by definition has the lowest possible score on inbound links, and it's unlikely to have any score boosting templates. The related page feature is extremely unlikely to select orphans as Read more links.

Indeed, right now, it is not how the feature works, but this could be something to explore and change, to help discoverability of more articles.

That looks like a reasonable plan. I would try to get most of the low hanging fruit tweaks done, before deploying the RFC on English (that might invite quite a range of IDONTLIKE responses, instead of constructive feedback responses).

@Sadads @Alsee This page shows the impact of removing popularity as a factor in the algorithm. What do you think? Extension:RelatedArticles/CirrusSearchComparison A lot less JK Rowling!!!