Microsoft Talk

Microsoft researcher Desney Tan led a team, including Greg Smith, Mary Czerwinski, Eric Horvitz and intern Michael Bernstein, to do just that. They developed a Facebook application called Collabio, a contraction of collaboration and biography, to explore a novel approach to gathering user meta-tags.

Trying to generate tags for people is not new. Because existing applications can use these tags in productive ways, many researchers have already been looking for how best to tag individual users. Systems have been built to automate the process by “scraping” email and Web pages. Another widely adopted method attempts to have people place tags on themselves.

But the results haven’t been stellar — Automated systems don’t work very well, and who’s going to write dozens of descriptive tags for him or herself?

Now, Facebook users can add the Collabio application and start slapping descriptive “tags” on their friends — lovely, Milwaukee, MIT, computer science, photography, mother, valedictorian, curly-haired, hamburgers, and so on. The tags range from geographic locations, to professional organizations and expertise, to hobbies and interests, to names of significant others or pets.

Collabio is designed to encourage small tags instead of sentences or paragraphs so that people make the hard choices, rather than relying on technology to distill the answers down. As such, having friends apply descriptive tags on one another is a simple approach to a problem that has been difficult to tackle. Collabio distributes that task to users on the network and makes it a fun social activity. Many participants in the game collect well over 100 tags. The highest is Bernstein, with 1,600.

In playing the game, users see a blanked out tag cloud for each friend and guess which tags have been applied. Each time two or more taggers agree on a tag, a little more of their tag cloud is revealed, and they get some points. Tags that have been placed on a person by multiple people are deemed to be more accurate. Scoreboards in the game keep track of who has the most points, tags and so on.

From a purely scientific standpoint, the game is a foray into exploring a new way to tap into knowledge contained within the social network and glean useful information for applications such as personalization or expert matching. Tan and team plan to analyze patterns of data to better understand the strengths and weaknesses of this approach.

The hope is that this novel approach will prove complementary to existing methods, and will inspire others to design useful applications that take advantage of the tag data. In the end, Desney Tan and his researchers believe technologies like Collabio have the potential to help turn social networks from somewhat of a novelty, into a useful and productive tool.

And for Collabio gamers themselves, an interesting side effect of the project is already emerging — the reflection they see of their friends’ perceptions. Users are widely reporting the experience as an interesting tool for personal introspection.

For example, one user who wears khakis to work every day reports seeing “khaki” pop up as a descriptive tag, and has started to contemplate his wardrobe.