ANN ARBOR—Internet marketers who try to predict what we like based on what we’ve bought online or which websites we’ve visited, may not know as much about us as they think do, according to new research at the University of Michigan.

In a study appearing in a forthcoming issue of the journal Social Psychological and Personality Science, U-M researchers Oscar Ybarra, Jeffrey Sanchez-Burks and Stephen Garcia find that people have an aversion to being predicted or labeled.

“With the help of modern technology, businesses are not only becoming more sophisticated in their attempts to figure out our preferences, they increasingly are showing us the results of their predictions,” said Sanchez-Burks, associate professor of management and organizations at U-M’s Ross School of Business.

“For example, Amazon.com tells me that I should buy a Kurt Vonnegut novel based on my preference for short stories by Calvino. TiVo first told an acquaintance he would like a series of racy films after he rented an episode of Sex in the City, but then suggested Neo-Nazi documentaries after he watched Gladiator. Delta predicts our colleague would enjoy a trip to Biarritz based on her recent visit to Istanbul.”

But the U-M researchers say that Amazon, TiVo, Delta and myriad other companies often miss the mark when trying to predict consumer preferences and behaviors.

In a pair of experiments with about 200 college students, the researchers found that people’s natural aversion to being predicted makes it difficult for companies to understand exactly what consumers want. Moreover, they argue that whether or not people want to be predicted depends on the interpersonal context, in particular, competitive or cooperative ones.

In a competitive context, people are more likely to go out of their way to be evasive, saying they don’t really have any particular preferences or values and that it depends on the situation, the researchers say. Even the mere suggestion that someone might be trying to predict us for some advantage may trigger a deep-seated response to be unpredictable?and we may not even be aware that we’ve behaved in this manner.

“Our findings say something about how potential consumers may react when a business feeds back their mathematical, magic eight-ball prediction of what we should buy based on what we watch, listen and buy,” Sanchez-Burks said. “Marketers be warned. Unless you are absolutely certain that your target audience sees you as a benevolent partner along their path to shopping nirvana, you might think twice about showing off your ability to predict consumer’s behavior.”

In addition to Ybarra, Sanchez-Burks and Garcia, other researchers who worked on the study include Matthew Keller of the University of Colorado, Emily Chan of Colorado College, Kimberly Rios Morrison of Ohio State University and Andrew Baron of Harvard University.