Sunday, February 24, 2008

Google Image Labeler: Letting the Community Do the Work For You

I did my Master's Thesis on ludological design -- the incorporation and application of video game elements into other media -- so it pleases me whenever I find similarly inspired work. I'm about a year and half late to this, but I recently discovered Google Image Labeler. The concept is deceptively simple and ingeniously effective.

When you sign in to play the game, you're automatically paired up with an anonymous partner. Over the span of two minutes, random images appear on screen and each player types in associating keywords. The object of the game is to come up with a matching word with your partner, thereby rewarding points to both players and bringing up another image. Meanwhile, on the side of the screen is a list of off-limit words that players cannot use, similar to the idea of Taboo. If you haven't figured it out yet, the game wasn't designed for pure entertainment; there is an ulterior motive...

You're doing work for Google.

The images are tagged with every keyword you match. And common matches end up being the off-limit words, thus preventing narrow results and permitting a wider range of tags. The Image Labeler is a great example of emergent design, creating a system for which the community produces the results that were once monotonously and unreliably inputted by users. And most importantly... it is fun.

You earn points for every matching word and your points are accumulated over time. At the end of each game, your score is compared to the all-time individual high scores, immediately giving you the incentive to try to reach the top. Constant display of high scores enforces the competitive nature of gamers and the short length of each game makes it so addicting. I frequently found myself saying, "I'll just do another one. It's only two more minutes." Before I knew it, I had played five more games. Google thrives on our competitiveness; the more games we play and the better we get at it, the better their image search becomes.

As gamers have learned from X-Box Live achievement points, a simple number representing their virtual accomplishments is something of value and is held in high regard by other members of the gaming community. It's amazing how something as arbitrary as an accumulative number, that originally didn't have any impact on the real world, can be so significant. Perhaps, gamers are so conditioned to equate worth to numerical figures as is common practice in role-playing games. Perhaps, it was always the hope that these virtual points would be transferable to utilitarian rewards. Whatever the reason is, the big companies didn't want to disappoint the ladder audience. Microsoft eventually turned that false hope into a reality, putting up items that players can spend their points on, and it's not a stretch to think that Google will go the same route with their Image Enabler points.

Increasing that self-representative number is fun, but the simple act of matching words is also incredibly engaging. Although the meta-game of getting the highest score is entirely competitive, the game itself relies solely on teamwork. In order to win, you have to think like your teammate. Since your teammate is randomly chosen and remains anonymous, this can be quite hard for the first image you both receive. But overcoming that first obstacle, you learn a little bit about your partner... he notices shapes first, she is very color-oriented, he associates abstract images of people to specific famous personalities, she prefers to point out verbs before nouns, he makes variations on the off-limit words, and so on. It's satisfying to get into your teammate's head, to form an ESP-like bond to someone you've never communicated with, to interact in a synergistic relationship... all within two-minutes.

By combining image tagging with a ludological structure, Google's created a system that produces intuitive and more reliable results. Individual tagging, on the other hand, is ultimately biased on the part of the user and can be easily abused. Not to mention that the Image Labeler is designed to encourage users to come back for more, iteratively improving the process along the way. Everyone else, catch up! I'm looking at you Flickr.

No comments: