Another HCI Blog

I lost one hour and thirty-eight minutes of my life last night when a user-friendly feature of the popular movie-rental-website, Netflix, backfired on me.

My partner and I sat down to enjoy some onion rings, snooty beer and the movie Poster Boy. I found myself, not 15 minutes later, visiting the Netflix website to a) figure out why I rented such a horrible movie; and b) save some sort of face in front of my partner who was relentless with her nose pinching P.U.’s.

I saw that it was recommended to me and rated an impressive four and a half stars. I notified my partner as much and we shrugged and kept watching.

“It must get better.”

It didn’t. And I soon found myself back at Netflix, searching for more justification. That’s when I discovered raters like me. Apparently, in what would seem to be a smart move, Netflix has grouped people based on their similarities and provides two rating averages: “raters like you” and “everybody.” When we logged on to my partner’s Netflix site, the movie rated a rental-discouraging two and a half stars.

On first glance this rating scheme sounds like a great idea. Netflix met a critical mass with their recommendation feature, and grouping together like-minded renters sounds like a sure fire solution for increased accuracy. And maybe that’s just what it does… except when it doesn’t. Last night when it didn’t, I was left with an entirely unexpected and unique problem: How does Netflix determine what kind of rater I am, and how can I fix it?

And then I remembered something.

In 2004, when I first started rating things on Netflix, I rated them assuming that I was smarter than the system. For example, I rated The L Word highly even though I wasn’t a big fan, figuring that would “teach” the system that I like gay-themed movies. In the beginning, I didn’t trust the system enough.

Years pass and I continue rating and renting and the system continues averaging and recommending. As I slowly and surely take those recommendations, I am rewarded. I begin to trust the system (maybe too much).

Now I’m left in the lurch. I don’t know how to fix my original (perhaps justified at the time, but clearly myopic) ratings. And any attempt to pretend I know the system now will surely result in a similar future fate.

The point is this: Historically an evangelist, I have created a problem all of my own volition. The result is that I’m unhappy with and distrustful of a service that was trying to help me out. These are times of growing pains for companies who arrive early at truly user-driven innovations.