My comment on Google Knowledge relates to a presentation in one of the WMF Metrics meetings and subsequent discussion (all of which can be seen on video since about 2012):    (Heather, this is a better URL than the one I sent you by accident a moment ago)


There are agendas but, while they tell you who will be speaking and about what topic, they don’t provide enough info for me to say which meeting was the one when they had discussed Google Knowledge as it was probably part of an agenda item labelled something like “top line metrics”.


I find it’s worth viewing these videos as they give the best insight into what’s happening at WMF at the moment. And of course there is always a metrics presentation, discussion of features in development, in beta, rolled out etc.





From: Heather Ford []
Sent: Wednesday, 11 June 2014 6:53 PM
To:; Research into Wikimedia content and communities
Subject: Re: [Wiki-research-l] Quality on different language version


Totally agree with you, Kerry - that there are *very* different ideas about what constitutes quality. The large diversity in research about quality taking very different variables into account is testament to that. I'm interested in your note about page views after Google Knowledge Graph. According to these stats Page views went up from mid-2012 to beginning of 2013 and then they went down quite sharply but seem to start rising again at the end of 2013. But perhaps you're seeing other data? Would love to hear your thoughts!


btw, for those asking about historiography, Brendan Luyt [1] has done some great work on how Wikipedia represents dominant and alternative historiographies [e.g. 2]. 


And thanks, Finn, for the great work you continue to do with Wikilit :)






On 11 June 2014 01:19, Kerry Raymond <> wrote:

Having followed this thread, I am somewhat confused about what is meant by
the term "article quality", even in a single language, yet alone multiple

Sticking just to a single language for the moment ...

Do we mean that the facts presented are correct? That the kings and queens
were born and died on the dates stated?

Do we mean spelling and grammar is correct? Do we mean some kind of logical
structure? Do we mean some kind of narrative flow that "tells the story" of
the topic in a natural and engaging way?

Do we mean the use of citations? Do we mean whether the citation used
actually contains information that supports what is said by the text in the
article with which it is associated?

Do we mean some kind of "completeness" of an article? That is, it has "all"
the information. If so, what do we do if the topic is split across a number
of articles {{main|...}}}? Do we assess the group of articles? And what do
we mean by "all" anyway?

Do we mean it meets all the WP policies? Notability? Appropriate use of
external links? That the Manual of Style has been carefully followed?

Or do we mean whether it has been assessed as a stub/start/.../good article
by some review process?

Whenever I find myself in a discussion about "quality" (on any subject, not
just Wikipedia), it pretty much always boils down to "fitness for purpose as
perceived by the user". This is why surveying of users is often used to
measure quality. "How well did we serve you today?" If anyone has been
through Singapore Airport recently, you will have encountered the touch
screens asking to rate on a 1-5 scale just about everything you could
imagine, every toilet block, every immigration queue, etc. And it does have
the cleanest toilets and the fastest immigration queues, so maybe there's
something to be said for the approach.

I think we need to have some common understanding of what we mean by
quality, before we try to compare it across languages. And when we do
compare across languages, then we have to observe that the set of users
changes and presumably their needs change too.

It is interesting to note that en.WP page views have dropped consistently
since Google Knowledge (which generally displays the first para from the
en.WP article) was introduced. What this tells us is that a certain
percentage of readers of an article simply want the most basic facts, which
would be delivered even by a stub article. "Suriname is a country on the
northeastern Atlantic coast of South America" certainly met my information
needs adequately (I heard it mentioned on the TV news in connection with a
hurricane). After finding out where it was in the world, I could have gone
on to read about its colonial history, its demographic sexuality, and its
biodiversity, but I didn't because I didn't have a need to know at that
moment. My point here is that while we would not generally regard a stub as
"quality", but a percentage of the readers of a stub are probably completely

Of course, doing surveys of articles with real users is somewhat difficult
for a research project. But it might be useful to see how user perceptions
of quality compare with other metrics (particularly those which can be more
easily generated for a research project). Starting with other metrics,
without knowing that they are a good proxy for user perception, is probably
a waste of time.


Wiki-research-l mailing list