Societal patterns and habits inherent in these data sets can predict behavior, such as the likelihood of residents of a certain neighborhood developing diabetes or alcoholism. (Predicting behavior from verbal and visual cues, it turns out, is rather easy; technology that Pentland and his team have developed is used by two large health insurers to screen callers for signs of depression.)
However, if exposure to external forces drives behavior changes, Pentland says, then getting to the root of the problem means changing exposure. Through its research, Pentland's lab reports that social influence—knowing that others are being rewarded for good behavior such as riding a bicycle to the office—is more than three times as effective as simply receiving that reward on an individual basis.