Fighting Recommendations

Many of the designers I know (or at least know of) have greedily downloaded a bunch of the TED conference talks. And while some reviews are mixed, most have been the sort of glow-y, OMFG, you have to listen to this! sort--the type of hype I feel like I've spent my whole life avoiding.

Why is this? And I don't mean, why am I a crank, but rather how come some things that we should like, we simply don't (or simply don't care to take the recommendation)? Amazon offers me up stuff all the time that I can't stand, despite having nearly a decade's worth of purchases to make recommendations from. Tivo, after several years of data, is hopeless when recommending television shows. Music has traditionally had the worst recommendations for me. Celebrated bands that I really should love leave me cold.

I don't think that I'm that particular--ok, I am--but not so much that a ton of data couldn't overcome it. Or is it that the nuance simply isn't there yet? That there aren't enough data points for entertainment content yet for predictions to be more accurate? Steven Johnson has a great example from one of his books: searching for a popular alternative Seattle band with crunchy guitars and a passionate lead singer from 1990-1994 will get me both Pearl Jam and Nirvana, and there's a significant difference between the two: to humans, anyway.

But even human recommendations can be faulty. Most of the people I know who like similar music to me love Neutral Milk Hotel and are always shocked when I'm indifferent. I'm the kind of pretentious jerk who would like the British version of "The Office" better than the US one, but actually it is the opposite.

As I've said in the past, recommendations don't take into account whim. I might download an Ashlee Simpson song ("Autobiography" say), but that doesn't mean I want all her songs. Sometimes off-beat stuff you'd never otherwise like, you do. For me lately, it's the ridiculously catchy song "Crazy" by Gnarles Barkley. "Crazy" would likely never be recommended to me, but I dig it nonetheless. How do we account for that?

Most of my "recommendations" lately come from unplanned encounters--flipping TV channels, radio, co-workers' networked music, books my wife has bought and I picked up. How do we--or should we?--design for that?

Originally posted at Monday, July 17, 2006 | Comments (1)