MovieLens
Can a web site recommend movies you would actually enjoy watching?
One web site claims it can do just that: MovieLens, a site sponsored by researchers at the University of Minnesota.
After you quickly create a free account, the site asks you to rate on a scale of one to five current releases and classic flicks. MovieLens matches your pattern of movie choices against the patterns of more than 50,000 other people who have created accounts at the site so far. The site then recommends that you see films rated highly by people who had a similar pattern of rating films that you did. Capsule plot descriptions are available through links to the Internet Movie Database.
I've found that I had to rate about 15 movies before I started getting useful recommendations. Rating fewer than 15 films did not give the site enough information to make useful comparisons. In the past twelve months, I've rated 100 movies, and the site's predictions have become uncannily accurate.
I created a second test account, randomly rating eight movies, to make plain how it works. I gave Amistad and The Boiler Room three stars each. I gave Saving Private Ryan, Sleepless in Seattle, Steel Magnolias, and The Straight Story top ratings of five stars each. And I gave Dumb & Dumber and Magnolia an awful rating of one star each.
Based on that pattern, MovieLens predicted I would like the following November 2000 cineplex releases: Solas, a well-acted Spanish film; Remember the Titans, a mushy, redemptive tale of high school football heroes; Urbania, which had the feel of Blair-Witch-Project -heads-to-the-gay-West Village-district-of-Manhattan, and lastly, the awfully formulaic Pay It Forward. The site also recommended DVD releases and older films on videotape that were liked by people who had given ratings for the eight films listed above that were similar to the ratings I had given them.
All this taste comparing is done anonymously, and none of your personally identifying information is shared with marketers. Better still, after a year of using my account I have not received any junk mail, despite having given my e-mail address when I opened my account. The site is a non-profit venture, and its popularity is ever increasing. It does not appear MovieLens will dry up and disappear any time soon. (Note of caution: The site's chief designer did not reply to my e-mail requests to confirm MovieLens's durability.)
One problem, though, is that the woo-woo fun of first using the site diminishes with time. After using MovieLens for a year, I have to admit I seldom go out to see a film in the theater based on one of its recommendations. If I'm stumped for movies to rent, though, and can't remember films I may have missed seeing in the cineplex, I sometimes make a list of MovieLens's predictions and bring it with me to the rental store. I've been pleased with the recommended films.
The biggest trouble with collaborative filtering technology, of which MovieLens is an example, is that it remains impersonal, whereas choosing art, such as a film, is very intimate. Ordinarily, I learn about a lot of films from reviewers who have a personality that comes through their prose and that I trust, or from friends who have a similar personality to my own. While technically MovieLens is duplicating that experience, it still seems cold. When the MIT Media Lab reportedly tried the same collaborative filtering with music recommendations at a site Firefly.com, not enough users were attracted.
MovieLens is easy to use, although if you have a slow dial-up connection to the Internet, rating movies can be slow. Most importantly, the recommendations are effective. But does getting movie picks from a machine subtract from the social fun of picking and choosing and evaluating movies in a culture of fellow film lovers? Only you can decide.
Sean O'Neill (NewsFromDC@cs.com)