问答题

Most of us tell one or two lies a day, according to scientists who study these things. And we rarely get caught, because the lies we tell are usually little ones: "I got stuck in traffic." "That color looks good on you." "I was just about to call."
But even the smallest fib may soon be systematically exposed, at least in the virtual world. Researchers at several universities are developing software that can detect lies in online communications such as instant messages e-mails and chatrooms. The ability to spot "digital deception", as researchers call it, has never been more crucial. Today, much of our business and social life is conducted online, making us increasingly vulnerable. White collar criminals, sexual predators, scammers, identity thieves and even terrorists surf the same Web as the rest of us.
Conventional lie detectors look for physiological signs of anxiety- a bead of sweat or a racing pulse--but online systems examine only the liar’s words. "When we’re looking at language, we’re looking at the tool of the lie," says Jeff Hancock, an assistant professor of communication and a member of the faculty of computing and information science at Cornell University.
Hancock, who recently received a $680,000 grant from the National Science Foundation to study digital deception, says there is a growing body of evidence that the language of dishonest messages is different than that of honest ones. For example, one study led by Hancock and due to be published this spring in Discourse Processes found that deceptive e-mail messages contained 28 percent more words on average and used a higher percentage of words associated with negative emotions than did truthful messages. Liars also tend to use fewer first-person references (such as the pronoun ’T’) and more third-person references (such as "he" and "they"). This may be the liar’s subconscious way of distancing himself from his lie.
More surprising, Hancock and his colleagues have observed that the targets of liars also exhibit distinctive language patterns. For instance, people who are being deceived often use shorter sentences and ask more questions. Even though they may not be aware that they are being lied to, people seem to exhibit subconscious suspicions.
To identify the patterns of deceit, Hancock has developed an instant-messaging system at Cornell that asks users to rate the deceptiveness of each message they send. The system has already collected 10,000 messages, of which about 6 percent qualify as patently deceptive. Eventually the results will be incorporated into software that analyzes incoming messages.
For now, the Cornell researchers are working only with the kinds of lies told by students and faculty. It remains to be seen whether such a system can be scaled up to handle "big" lies, such as messages sent by con artists and terrorists.
Fortunately, the research so far suggests that people lie less often in e-mail than face-to-face or on the phone. Perhaps this is because people are reluctant to put their lies in writing, Hancock speculates. "An e-mail generates multiple copies," he says. "It will last longer than something carved in rock." So choose your words carefully. The Intemet may soon be rid not only deceit but also of lame excuses.
Why does Hancock need the rating results of the message deceptiveness