Gaming for Google

Oct 6
2006

A couple days ago I came across Google’s Image Labeler.

Essentially, the Image Labeler is a system to allow users to tag images with keywords, with the goal of providing more accurate image search results. The genius of Google’s method is that rather than trying to convince users to give up some of their time to sift through and tag millions of images, they turn it into a game.

Here’s how it works:
You are paired randomly with another user. The Image Labeler presents you and your partner with an image from Google Images. You now both begin suggesting labels for this image, without knowing what labels your partner is suggesting. If both of you suggest the same label you move on to the next image. The goal of this ‘game’ is to get through as many images as possible in 90 seconds. You are awarded 100 points per image and a scoreboard shows how well you did at the end of the round.

I wouldn’t exactly say this game is fun, but it is strangely addicting. It’s quite challenging to think of labels for some images, and it’s interesting to see how well you perform with certain partners compared to others. Sometimes I feel like I really ‘click’ with someone and we’re able to get through 9 or 10 images. Other times my partner and I can’t even get one match.

What’s really impressive about the Image Labeler is that Google has made an interesting game out of, well, work. Not only do you provide them labels free of charge, but by structuring it the way they have, they will probably end up with better keywords than they ever would have if they had hired a group of people to provide the same service. They are getting input from two different people on each image. And assuming that the users care about scoring points in the game, each user has a motivation to provide the best keywords he can, to increase the chances of matching one of his partner’s.

Now if only I can come up with a game that will trick people into doing my work for me…