A Modest Proposal: The Netflix Jury

I received a questionnaire for jury duty yesterday in the mail. It wasnâ''t a summons, though surely it will lead to that. I donâ''t mind. Serving on a jury is one of our few civic duties, a cornerstone of free and fair trials, which itself is a cornerstone of democracy.

The notice says that my name was culled at random from voter registration, driver registration, unemployment, or other social service records. I have no problem with that. But it did make me stop and think. Thatâ''s not a bad way to come up with a jury of my peers â'' I do, after all, vote, drive, and rely on the social safety net from time to time â'' but to really come up with a jury of my peers, how about getting records from Netflix?

Hereâ''s what I have in mind. Netflix already has a system for comparing my movie ratings and the films in my queue to the ratings and queues of every other Netflix subscriber. Itâ''s the basis of, among other things, Netflixâ''s recommendations â'' the feature by which it says, â''Viewers who liked this movie also liked....â''

So in my imagined system, Netflix sends the city a list of every other Netflix subscriber in nearby zipcodes who matches my ratings and queue above a particular threshold. Now that would be a jury of my peers!

Of course, movie preferences track with citizenship only imperfectly. Even among my friends, who I as a defendant would love to see on a jury, there are some with just terrible taste in movies, that is, they disagree with me as to what the best movies are. And, no doubt, plenty of people who share my cinematic tastes and yet would put me behind bars without even listening to the evidence.

So a better system might be to find people who read the news in the same way I do.

Imagine a Website that lets you read current news stories. Every time you click on a headline, the software makes a small note of it. It quickly begins to compile a profile of your newsreading preferences based on these notes and recommends other news stories to you, based on the behavior other people reading news at the site and the profiles it has compiled of them. It then notices whether you click on the recommended stories or not, and so on.

Such a site existed from 2004 to 2007. Called Findory, it was developed by Greg Linden, the author of â''People who Read This Article Also Read....â'' in this monthâ''s issue of Spectrum. Linden wrote Amazon's original recommendation system.

Greg doesnâ''t speculate on the jury-building possibilities of such software, but he does consider its potential for reviving newspapers as that industry moves inexorably online. He describes the thorny technical challenges he and researchers at Google and elsewhere have encountered in designing and applying recommender-software.

Sure, there are grave privacy issues with the idea of Netflix or Findory (or Google or the New York Times) handing our personal information over to the city or any other government. But at least weâ''d know who hereabouts has newsreading habits and preferences that are similar to mine. If I canâ''t have them on my jury, maybe I can at least meet them for dinner â'' and a movie?


Tech Talk

IEEE Spectrum’s general technology blog, featuring news, analysis, and opinions about engineering, consumer electronics, and technology and society, from the editorial staff and freelance contributors.

Newsletter Sign Up

Sign up for the Tech Alert newsletter and receive ground-breaking technology and science news from IEEE Spectrum every Thursday.