Games with a purpose

Computer games are often viewed as solitary and addictive. But for Luis von Ahn, a 29-year-old computer science professor at Carnegie Mellon University, they can be strategic, communal – and useful. 

The winner of a MacArthur “genius grant,” Ahn is a pioneer in developing games that channel people into working on tasks – such as indexing vast amounts of communal data – that computers are unable to handle. While tapping into human tendencies toward procrastination or entertainment, Ahn has been able to create games that produce data that can applied to a number of tasks that can exceed the ability of computers, such as translating documents accurately or improving search-tool accuracy.
 
In an Oct. 8 lecture, Ahn talked with more than 60 Princeton undergraduates, graduate students, and faculty in the Computer Science Building about his groundbreaking work on the “ESP game.” Since computers lack the ability to recognize images, this game encourages humans to label them; the labels cumulatively build up great quantities of accurate data. 

During the game, a player is paired up with an anonymous partner. Separately, they are shown the same image and required to describe it in a word. Points are awarded when there is mutual agreement on a word, which ensures the accuracy of the ultimate “label” that is recorded. “It’s possible to really feel a connection with your partner,” Ahn said, “especially when you agree on an off-the-wall word like ‘yuck’ for President Bush.”

This connection, of course, might be challenged by an individual’s ambition to “win” the game. Ahn commented dryly: “If you do well, you fall in love with the partner. If not, they’re an idiot … of course, they’re the idiot.” Students asked Ahn about how a competitive nature might affect such games, as well as the potential for lucrative business applications. 

Yet Ahn’s vision stems from a desire to allow computer users to change the world for the better. His concepts can be applied to broader tasks that range from digitizing library books and archived materials to improving the accuracy of security X-ray screenings at airports. His dream, he said, is that someday computers might come close to “learning” human skills themselves, recognizing faces, speech, and handwriting – even distinguishing the artistic qualities of languages and images. 

At least one student was skeptical about the concept, however. “As I understand it, to procrastinate real work, humans play games that actual computers will always find impossible, simultaneously indexing mountains of data that describe the world,” said Chris Rucinski ’09, a computer science major. “I suppose this is why robots will never enslave humanity.”