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

[Gregory Maxwell]

On Wed, Jun 2, 2010 at 3:48 PM, Aryeh Gregor
<Simetrical+wikilist at gmail.com> wrote:
> On Wed, Jun 2, 2010 at 2:50 PM, Gregory Maxwell <gmaxwell at gmail.com> wrote:
>> Who cares if people click them a lot?  The space they formally
>> occupied is filled with nothing now.
>
> Interface clutter is not psychologically free.  Empty space is better
> than space filled with mostly-useless controls.  Whether these
> particular controls are worth it I don't know, but the general
> principle of hiding seldom-used things is sound.


Mostly-useless is not the same as infrequent.

I think that this is a critical error people make when trying to be
data-driven. (I strongly support and promote data driven decision
making, but also fear that it can be so frequently misused)

For example, the stats show 14 million page views a day from
blackberry.  14m is not frequent compared to 3.6billion.  (0.0038 BB
users per user), and yet that number represents an enormous number of
people in _absolute terms_ whos ability to use Wikipedia may be
degraded, disrupted, or completely inhibited right now.

Likewise, I do not doubt that only "One in big_number" followed an
interwiki link, but the fact that more people use the site doesn't
remove the value received by the smaller group ... which, because of
the size of the site, might still consist of a hundred thousand
people.


You can attempt a weighted cost comparison:    Num_interwiki_users *
Cost_of_hiding   vs   Everyone_else * Cost_of_clutter.    But even
that will inevitably lead to bad conclusions for some issues because
the costs are usually not linear things:   A tiny benefit to a hundred
million people wouldn't justify making wikipedia very hard to use for
a hundred thousand,  ... because a zillion tiny benefits can often
never really offset a smaller number of big costs.

Contention about applying linear costing to things with non-linear or
outright incomparable values is why people get worked up when deaths
get imported into a cost/benefit analysis.   It is somewhat crude of
me to compare killing someone with inhibiting their use of Wikipedia,
but I think the same problem with cost/benefit analysis exists there—
if only in a very reduced form.