A Brigham Young University study suggests people when people lie in digital messages (texts, social media or instant messages), they take longer to respond, make more edits and write shorter sentences than usual.
"Digital conversations are a fertile ground for deception because people can easily conceal their identity and their messages often appear credible," says Tom Meservy, BYU professor of information systems. "Unfortunately, humans are terrible at detecting deception. We're creating methods to correct that."
According to Meservy, humans can detect lies about 54 percent of the time accurately -- not much better than a coin flip. It's even harder to tell when someone is lying through a digital message because you can't hear a voice or see an expression.
With the many financial, security and personal safety implications of digital deception, Meservy and fellow BYU professor Jeffrey Jenkins, along with colleagues at the University of Nebraska-Omaha and the University of Arizona, set up an experimental instrument that tracked possible cues of online lying.
The researchers created a computer program that carried out online conversations with participants -- similar to the experience consumers have with online customer service questions.
More than 100 students from two large universities, one in the southeastern U.S. and one in the southwestern U.S., had conversations with the computer, which asked them 30 questions each.
The participants were told to lie in about half of their responses. The researchers found responses filled with lies took 10 percent longer to create and were edited more than truthful messages.
"We are starting to identify signs given off by individuals that aren't easily tracked by humans," Meservy said. "The potential is that chat-based systems could be created to track deception in real-time."
The findings appear online this week in the academic information systems journal ACM Transactions on Management Information Systems.
Meservy and Jenkins, who coauthored the study, said we shouldn't automatically assume someone is lying if they take longer to respond, but the study does provide some general patterns.