Online Psych Test Catches Conflict In Undecided Voters

This story was written by Evan T. R. Rosenman, Harvard Crimson

About 16 percent of voters still haven't made up their minds in the presidential election, according to researchers at Harvard, the University ofVirginia and the University ofWashington.

The three scientists are collaborating on Project Implicit, a research Web site that allows visitors to complete various tests gauging their subconscious associations. The tests cover a wide variety of topics, including racial, religious and gender biases, as well as preferences among the presidential candidates.

As psychologists, what were doing is using the Web site to understand some of the mechanisms that go into the act of voting how our attitudes and preferences about social groups might affect our choice of a candidate, said Harvard psychology professor Mahzarin R. Banaji.

More than 25,000 participants took the Obama-McCain test. Of those, 16 percent --about 4,000 -- claimed to be undecided.

The results revealed that many of the undecided voters demonstrated an implicit preference for one candidate or the other. On average, participants reported feeling slightly greater affinity for Obama, but their implicit biases leaned toward McCain.

They have this conflict, said UVa researcher Brian Nosek. Explicitly they want to like Obama, but implicitly they like McCain. Perhaps thats playing a role in their undecided status.

The Implicit Association Test Web site recently celebrated its 10th anniversary, meaning that this is the third election cycle for which tests have been conducted. In 2000, the studies were surprisingly accurate predictors of primary winners. As Banaji recalled, many Democrats were explicitly supporting Bill Bradley, while many Republicans claimed to be supporting McCain. But results from the test showed that the same voters were implicitly biased toward Al Gore 69 and George W. Bush, respectively the two eventual nominees.

But the researchers cautioned against using their data to forecast the outcome of next Tuesdays election. While Nosek said that the undecided voters may break slightly more for McCain, he added that he expected the overall effect to be relatively small.

Banaji also noted that because the data was collected online, it inherently underestimates the poor, those who are not computer literate, and perhaps even conservatives.

Nonetheless, the research presents fascinating questions for the academic world. Harvard Psychology Professor Ellen J. Langer called the work fabulously important. She added, I think that our implicit biases can affect virtually all that we do. The importance of the work is pervasive.