Wednesday, May 30, 2012

Advertising Algorithm FAIL

Today was perhaps the worst of a series of bad days in Seattle. Gun violence here has been incredible.

A man suspected of shooting five people at a University District cafe — and a woman later on First Hill — shot himself as police closed in.

Turns out today's maniac — mentally ill of course, god forbid we hospitalize these folks — once lived next door to my son Nick, who called me this afternoon to mention that the SWAT team was in his back yard and cops were busting into the neighbors' place. (One of the shootings took place in a coffee shop down the street which Nick frequents.)

What has this to do with UI text, you might ask?

Well, see, I spent a fair amount of time on Facebook today monitoring the situation and chatting back and forth with friends, family, and Nick.

And here, several hours after the fact, FB blithely displays for me a bunch of Recommended Pages in the sidebar, including "Books" (3 of my friends like) and "Shooting" (1,668,817 people like).

Aside from that nauseating statistic, why the frack does Facebook want to recommend THAT to ME, TODAY of all days?

Oh, and when I delete that one, I get "Guns" (1,987,504 likes). I guess I should be grateful that there are 318,687 weirdos on Facebook who just like guns but not shooting them.

And my point is?

I figure it has to be that FB is monitoring my posts, or the posts I recently commented on, and finding the common denominator. Which in the case of chatting about movies or music or cars or Twinkies or whatever would be insidious enough, but where their algorithm FAILS is not deducing that I'm posting about guns and shooting not because I like them but because I think they're an abomination.

Monday, May 28, 2012

Icons That No Longer Make Sense

Must be a sign of harmonic resonance or something. I was just talking about this sort of thing with my 14-year-old daughter (and she was the one who brought it up!) the other day.
Icons That Don't make Sense Anymore

By the way, Blogger, since when do I use a pencil to create a new blog post?

Thursday, May 03, 2012

The perils of appearing accurate

At work the other day, we were looking at a prototype of some software. The concern arose that, when we unveiled the prototype to the PTB (Powers That Be), they might get the wrong impression: if it looked too finished, they would take the demo UI too literally. At the very least the PTB could begin unnecessarily nitpicking... "rat-holing," in the common parlance.

Oddly, this misapprehension can occur with released products too. They look so darn good, so authoritative. In fact, most products out there will have some flaws, but the trick is to not let the consumer see them. Naturally this requires a bit of smoke and a couple of mirrors.

But the example of deceptive accuracy that sprang to mind as I was looking at our prototype was a set of IKEA assembly instructions. Now, by and large these are pretty decent, not least because they use little or no language, just illustrations, hence you get none of the infamous "Engrish". But the illustrations look very convincing. The number of screws is listed, there are helpful dotted lines showing where they go, and so on. They are of course too small to be really super-clear, but they do look pretty dang painstaking.

The key syllable, though, is "pain." Seldom have I assembled an IKEA item without at some point having to undo some or all of my work and attack it again — not necessarily because of human error (mine) but because some vital morsel of information has been left out, or is unclear. A screw may not have been drawn correctly. The number of screw holes drawn may not match the number of screws in my packet. In short, the seeming accuracy, the spiffy “fit and finish” product feature (in this case the manual rather than the object to be built) fools me into trusting it.

Now, there’s always the chance that I’m simply so anal that I fixate on the “perfection” or lack thereof, rather than just using my common sense and building the darn bookcase using the Occam’s Razor approach. But perhaps in some cases it is better to be vaguer in the implementation: "describing rather than prescribing" might be a way to think about it.