»  E-Mail »  Print »  Discuss »  Del.icio.us »  Digg

I was at a wedding once (it might've been mine), and the preacher talked about how a lot of people say marriage is hard. You've really got to work at marriage, people say.

"Marriage isn't hard," the preacher said. "But you do have to pay attention."

It's good advice--listen to your spouse, understand the nuances, and you won't be caught off guard when a huge fight erupts over your nightly garlic pork rind bedtime snack. Pay attention when someone's sending you a message.

Which is why I don't entirely buy all concerns raised about a $2.4 million Department of Homeland Security grant to help the United States pay attention to what's being said about it. The department is sponsoring computer science researchers at universities in hopes they'll develop a system that automatically makes judgments about articles in foreign newspapers and other publications, such as gauging whether they hold positive or negative opinions, and how strong the feelings are.

Lucy Dalglish, executive director of the Reporters Committee for Freedom of the Press, called it "creepy and Orwellian" in an article in The New York Times, which wrote about the project last week. Understand, since Dalglish's group works full time on issues like keeping journalists out of jail, there's nothing not to like about her. But I couldn't figure why the government shouldn't use tools to track what people are publicly saying about it. "If they can pull it off, it's quite an intellectual and technical achievement," Dalglish told me. "But there's a part of me that just thinks we need human beings to make judgments on things like opinion." Could such a system really figure out sarcasm, hyperbole, and tone? she wonders.

The technology does sound devilishly hard to do. The project needs to create "trainable learning algorithms" that can sort subjective comments from fact, identify beliefs held by a person or group, and "assess the intensity, polarity, and motivation and attitude types of those beliefs," according to a description of the research on the Web site of the University of Pittsburgh, where computer science professor Janyce Wiebe is among the project researchers. A fellow researcher, Cornell computer science professor Claire Cardie, describes in a university publication how learning systems can pick up triggers that might not be intuitive, such as that the words "it's a fact that ..." are often prelude to someone stating an opinion.

The results would include links that let people follow through to the source material to check the software's conclusion. But all that's just the technical part. Then there's the policy side. "There are people in the government I don't trust to use it appropriately," Dalglish says.

We have reason to fear abuse of technology. The government promises not to use it for U.S. publications. But if these systems get good, there will be tough questions: Should foreign-based blogs be fair game? Chat rooms? Are those closer to publications or coffeehouses when it comes to whether casual surveillance and analysis is appropriate? OK, you can start to see "creepy" peek over the horizon.

But the United States is a country--its leaders and general population--famous for not paying attention to the world. So I wouldn't mind an algorithm or two keeping an ear out for us. Ultimately, if there are tools to help us pay attention, use them properly. But use them. Best of luck to the research team.

If John Soat were writing this column, he could've cooked up a conspiracy theory about using this to control our thoughts. He'll return next week. Contact me at cjmurphy@cmp.com, or welcome him back with a tip at jsoat@cmp.com, or by phone at 516-562-5326.

To find out more about Chris Murphy, please visit his page.

	»  E-Mail »  Print »  Discuss »  Del.icio.us »  Digg »  MyInformationWeek »  Reprint This Article »  Download Top Reports