Concepts familiar from grade-school algebra have broad ramifications in computer science.
What you say in a conversation -- whether it's on a first date, a job interview or pitching an idea -- may be less important than how you say it. But the cues that may decide the outcome can be so subtle that neither person in the conversation is consciously aware of them.
Whether or not you get the job, or the other person's phone number, is very strongly influenced by unconscious factors such as the way one person's speech patterns match the other's, the level of physical activity as people talk, and the degree to which one person sets the tone -- literally -- of the conversation. These subtle cues provide "honest signals" about what's really going on and strongly predict the outcome, according to research by the MIT Media Lab's Alex (Sandy) Pentland and his colleagues.
"Honest Signals" is also the title of Pentland's new book about the research, being published this month by MIT Press. The research was based on tens of thousands of hours of data from devices about the size of a credit card that record movements and voices, which Pentland has dubbed "sociometers." Using just this data, with no knowledge of what was said, Pentland could predict the outcome -- whether a job offer, a second date, or investment in a business plan -- more accurately than by using any other single factor.
Pentland says that this technology is recording and quantifying something that most people already understand intuitively. "All of this is sort of folk knowledge," he says, "we all know it's there, but we all ignore it."
Pentland, with both a degree in psychology and experience in signal processing, zeroed in on "a few things that seem to come up again and again" in deciding what aspects of human communication to monitor with the new devices.
The features he found that are highly predictive of outcomes, he says, "match the literature in biology about signaling in animals." In fact, Pentland suggests, the non-linguistic channels of communication that are measured by the sociometers may have started among our ancestors long before the evolution of language itself, forming a deeper, more primal way of understanding intentions, coordinating activities and establishing power relationships within the group.
"Half of our decision-making seems to be predicted by this unconscious channel," says Pentland, the Toshiba Professor of Media Arts and Sciences. "That's exactly the channel that you see in apes" as they coordinate their activities without the use of language.
Pentland's research so far on these non-linguistic signaling channels has been based on getting groups of people, such as attendees at a conference or employees of a company, to wear the sociometers over periods ranging from a day to a month or more. The devices, which include a microphone for recording voices and accelerometers to measure a person's movements, are a bit smaller than the name badges typically worn at conferences. In future research, he says, the same functions could be monitored using specially programmed cellphones.
The data gathered from the devices can be used not only to predict the outcomes of specific interactions between people, but even the relative productivity of different teams within a company. "This information is not in the organizational charts," Pentland says. "This human side is missing from all traditional measures" of how groups of people work together.
The strong correlations between unconscious forms of communication and the decisions that result strongly undermines people's perception that they are making choices based on rational, conscious factors, Pentland says. "My data shows that's simply not true." By understanding and measuring factors that people are usually unaware of, he says, "I view it as putting human nature back into our social fabric."
It may even help to predict the outcome of elections, he says. For example, by watching for the movements that signal the factor Pentland calls "influence" -- the setting by one person of the tone and pace of a conversation -- in a presidential debate, it is possible to see which person is dominant, regardless of what is being said. "The person who sets the tone," he says, "is the one who wins, in every election since 1960."