UW professors create, and consider, neuroscience technology that measures voters’ attitudes

Presidential polls may have met their match. New technology can read voters’ minds to see what they’re thinking — not just what they’re saying — about presidential candidates.

Although neuroscience may provide an explanation for surprising voter trends in the primary election, the validity of such technology is debatable.

“I look at the work [of neuroscientists] with a little criticism,” said Bethany Albertson, a UW assistant political science professor. “There are multiple interpretations for brain imaging, but applying it to the study of politics is very new.”

Albertson is referring to brain imaging technology most commonly used to measure cognitive and physiological responses to video games and commercial advertising.

Neuro-marketing firms nationwide are now pitching their services to presidential campaigns. EmSense Corporation has measured reactions to debates and campaign ads in New Hampshire and Iowa.

Brad Feldman, head political analyst for EmSense, said tests are administered to a group of at least 100 participants, who watch a campaign ad together in real time while wearing individual headsets.

“[The headset] measures brain waves and other biometric sensors like blinking, breathing and body temperature, and combines the data through algorithms and EEG technology,” Feldman said. “We can tell how much people are seeing [in an ad] and if they like it.”

Feldman believes neuroscience isn’t manipulative, but a way to access as much information as possible.

“When you think about elections, people aren’t entirely rational,” he said. “If you only have the conscious analysis, you are going to miss the emotional aspect.”

Yet Darren Schreiber, a UC San Diego assistant professor of political science who has studied brain imaging for the past 10 years, said: “We are certainly not at a level yet where we can interpret what the data means.”

“I interpret it very cautiously,” said Albertson, who distinguishes between brain imaging technology, such as the EmSense headset and the Implicit Attitude Test (IAT), a less invasive way to tap political attitudes.

Developed by UW psychology professor Anthony Greenwald, IAT is composed of a series of matching exercises on the Internet.

“IAT finds out about what’s in people’s heads that they may not tell you because they don’t have conscious access to it,” Greenwald said.

Unlike brain imaging, IAT has been used to measure subconscious voter attitudes for the last decade.

It is receiving more attention now because, unlike the 2000 and 2004 elections, Greenwald said tests administered in October and November, 2007 produced an inconsistency between what voters were saying in their self-report and what they were implicitly thinking.

A demo IAT, projectimplicit.net, reveals how such an inconsistency can come about. After selecting one’s preferred candidate in a self-report, participants are instructed to press one key when a positive word, such as “happy,” is presented with a picture of a particular candidate. Negative words are matched with another candidate. Then the association is reversed.

The test measures the accuracy and speed at which a participant is able to complete each matching task, believing that faster responses indicate stronger feelings for a particular candidate.

In recent tests, lots of subjects leaning toward Sen. Barack Obama in their self-report showed stronger support for Sen. Hilary Clinton on the IAT, Greenwald said.

“There seemed to be a problem with self-report,” he said. “We expected polls might be off because they are based on self-report.”

“It could be that people are flat out lying in polls, but I don’t think that’s what’s going on,” Albertson said. “I think they also have some negative bias on an unconscious level.”

Greenwald’s research confirms the presence of the “Bradley Effect,” a phenomenon named after the 1982 California governor candidate, Tom Bradley, a black Democrat who lost even though polls ranked him ahead of his white opponent.

Overestimated support for Obama in the New Hampshire primary and underestimated support in four other states, including Washington state, suggest that the “Bradley Effect” is a factor in this election, Greenwald said.

His findings show a correlation between the support for a black candidate and the proportion of black voters in a particular state.

“Years of implicit research show that Americans have a preference for white over black faces,” Albertson said. “Racial stereotypes might run true.”

In other words, IAT may reveal subconscious preferences based on more than emotion; it may suggest that people have hidden racist tendencies that factor into one’s decision to vote for a particular candidate.

“My biggest worry about the bias to Obama is that biases make us more vulnerable to negative information,” Albertson said.

But people might not necessarily act on these attitudes.

“The good news is we can override our impulses,” Schreiber said. “If someone hit your knee, you would kick your leg because it’s a reflex. But, you may have an impulse and decide not to do it.”

Please read our Comment policy.