[Foundation-l] hiding interlanguage links by default is a Bad Idea, part 2

[Aryeh Gregor]

On Fri, Jun 4, 2010 at 4:21 PM, David Levy <lifeisunfair at gmail.com> wrote:
> At the English Wikipedia, this is not so.  If we had a bike shed,
> there would be daily complaints about its color.

I should say that *almost no* users complain about small things.  A
tiny group of committed users will complain about small things, but
they're not the targets of the Usability Initiative, so their
complaints are not relevant here, *except* insofar as they provide
reasoning or evidence about what most users think.  By contrast,
complaints from occasional users are useful in usability discussions
even if the users provide no reasoning, because the complaints are
ipso facto evidence of a problem.  (But if we have only anecdotal
evidence of complaints from occasional users, of course, that needs to
be treated with the same caution as any anecdotal evidence.)

> I've encountered many complaints about clutter at the English
> Wikipedia (pertaining to articles, our main page and other pages), but
> not one complaint that the interwiki links caused clutter.

My first guess would be that people didn't complain about interwiki
links' clutter because they've always been there.  By the time you're
comfortable enough with the site to complain, you just won't notice
them.  I'd guess that the complaints you see are when things *change*.
 Experienced users are prone to complain when things change, because
they've gotten used to how things are.  If we leave off the links for
a year, then turn them back on, I predict we'd get complaints about
clutter.

> However, assuming that the interwiki links benefit a relatively small
> percentage of users (still a non-negligible number in absolute terms),
> I've yet to see evidence that displaying them by default is
> problematic.  Like David Gerard, I desire access to the data behind
> this decision.

Then say exactly what evidence you desire.  What test would you
suggest to see whether hiding the links helped or harmed things?

On Sat, Jun 5, 2010 at 1:30 AM, David Levy <lifeisunfair at gmail.com> wrote:
> Said data indicated only that the interwiki links were used relatively
> infrequently.  Apparently, there is absolutely no data suggesting that
> the full list's display posed a problem.  Rather, this is a hunch
> based upon the application of a general design principle whose
> relevance has not been established.
>
> Aryeh Gregor: You cited the importance of data (and the systematic
> analysis thereof).  In light of this explanation, what is your opinion
> now?

Data is important.  It's also not always possible to gather.  When
multiple things are competing for attention, you can make one or the
other more prominent, and it will get correspondingly more clicks.
But it's up to your judgment to assess whether that's a good thing or
a bad thing: are more people finding what they actually want, or are
people being distracted from what they actually want?  If we have more
clicks on interlanguage links and less on other interface elements, is
that good or bad?  If we wanted to maximize clicks on interlanguage
links, we could always put them above the article text, so you have to
scroll through them to get to the article text . . . but that's
obviously ridiculous.

As Greg said above, data is important, but it can be hard to apply
correctly.  Sometimes you really have to use judgment.  But we could
still use more data -- for instance, why do people usually click
interlanguage links?  Do they usually understand the language they're
reading the article in, or not?  We could have a little
multiple-choice question that pops up a small percentage of the time
when people click on an interlanguage link.

My suspicion is that a long list is not ideal.  Yes, people will see
it for what it is and they'll be able to find their language easily
enough if they look.  But it's distracting, and it's not obvious
without (in some cases) a lot of scrolling whether there's anything
below it.  If we could use some heuristic to pick a few languages to
display, with a prominent "More" link at the bottom, I suspect that
would be superior.

But first we should gather data on click rates for the list fully
expanded and unexpanded.  Per-page click rates are important here --
many articles have no interlanguage links, so will obviously pull down
the average click rate despite being unaffected by the change.  What's
the trend like as articles have more interlanguage links?  How many
more interlanguage clicks are there for articles in twenty languages
as opposed to five?  Can we plot that?  For each wiki separately, for
preference?

All this data gathering takes manpower to do, of course.  Maybe the
usability team doesn't have the manpower.  If so, does anyone
qualified volunteer?  If not, we have to make decisions without data
-- and that doesn't automatically mean "keep the status quo", nor
"change it back if people complain loudly".  It means someone who
happens to be in charge of making the decision needs to make a
judgment call, based on all the evidence they have available.

(By the way, I'm not an employee of Wikimedia and am sometimes not at
all happy with how the usability team operates.  I happen to think
that they have a good point in this case, though, irrespective of how
they made or enforced the decision.)