Research:External Reuse of Wikimedia Content/Wikidata Transclusion/Examples
This page will be used to document the many examples of Wikidata transclusion in Wikipedia and begin to categorize them based on a number of different criteria. Each criterion attempts to capture how the transclusion might affect readers or editors. With the exception perhaps of platform, each criterion below is quite subjective (though can probably be measured to a certain degree). Given the highly contextual nature of this work, I am sure that others would come up with different criteria (or interpretations of how to apply the existing criteria).
Does this type of Wikidata transclusion show up in all form factors or just desktop, mobile, iOS app, Android app, or some combination? There seem to be three main categories:
- All platforms: many types of content transclusion show up on all form factors (though placement may vary) -- e.g., infobox content, external links.
- Apps only: Wikidata descriptions are used only by the mobile apps as part of the search experience. This usage is not actually tracked as transclusion in the wbc_entity_usage table.
- Desktop only: Metadata tables like Authority Control and Taxonbar only show up on desktop. Currently, categories only show up on desktop as well.
- Desktop and apps only (not mobile): Coordinates when they are set to display in the title only show up on desktop (though can power a "View on Map" option in the apps).
Where in the page is the content? Note that content can be highly salient (when it appears) but only show up on certain platforms. We see three main categories:
- Highly salient: this includes content within the lead or infobox that will show up at the top of the page and is easy to distinguish. This also includes Wikidata descriptions, which are used in search results and are highly salient within that specific context.
- Medium salience: transclusion that affects a section of the article (e.g., adding external links) can be highly salient in stub articles but are buried in larger articles. Perhaps better referred to as central but highly variable salience.
- Low salience: for example, metadata templates or categories which will be at the bottom of the page and are set aside in a way that they are easy to skip.
How likely is the transclusion to affect a reader (especially if the content is incorrect)? This is perhaps the most subjective of the criteria (and there will certainly be plenty of counterexamples), but I do my best below:
- High impact: readily-interpretable facts, especially those about people, such as a birthdate. This sort of information is less likely to be fact-checked as well -- i.e. the reader does not have to follow a link or read much to get the information so it might be less obvious when it's wrong.
- Medium impact: links that are likely to be followed, especially when there is not much content on the page, and that could be easily manipulated to point to the wrong page.
- Low impact: categories, which are mainly relevant to maintenance, specifics of a reference, and metadata templates such as Authority Control or Taxonbar that have more limitations about the links that appear, are rarely clicked, and, when they are clicked, should be readily obvious if they are wrong and minimally offensive as the websites they lead to both provide plenty of context and are well-curated.
Bringing together these three criteria, we can categorize various types of Wikidata transclusion based on how important it is:
- Content in infoboxes shows up on all platforms, is highly salient, and can potentially be very impactful to readers.
- Wikidata descriptions only show up within Search and the Related Pages carousel on mobile apps currently, but they are easy to manipulate and highly salient to readers when visible.
- Commons category templates and external link templates such as IMDb title often are insignificant but can potentially be important for stub articles and could be manipulated to generate misinformation or troll individuals.
- Authority control and taxonbar only show up on desktop, have low salience, and the cost of disinformation is low.
- Tracking categories don't show up on most platforms and are low salience and impact when they do.
See this analysis of 100 random English Wikipedia articles.
- Piccardi, Tiziano; Redi, Miriam; Colavizza, Giovanni; West, Robert (26 January 2020). "Quantifying Engagement with Citations on Wikipedia". arXiv:2001.08614 [cs]. Retrieved 30 July 2020.
- Mitrevski, Blagoj; Piccardi, Tiziano; West, Robert (21 April 2020). "WikiHist.html: English Wikipedia's Full Revision History in HTML Format". arXiv:2001.10256 [cs]. Retrieved 30 July 2020.