Fact-Checking Google’s New Study On Political Fact-Checking: Mostly Misleading

Next Story

Samsung’s Galaxy Nexus Gets A Reprieve As Appeals Court Overturns Sales Injunction

Welcome to TechCrunch’s ongoing series where we hold major tech companies accountable for spreading propaganda about their service’s value to the democratic process. Today, Google released a “study” showing that swing voters are turning to the Internet in droves (64%) to fact-check candidates. We rate this study “mostly misleading” for a few reasons. 1). most “swing” voters are actually closeted partisans, and the most recent evidence on genuinely independent thinkers finds that they care more about personality than substance. 2). Most voters look for evidence confirming their pre-existing beliefs. 3). Surveys are an awful, no-good way of measuring behavior. But, on the other hand, experimental evidence finds that refusing to refute false statements is detrimental to campaigns, so fact-checking does have its benefits.

Google released a study by Global Strategy Group and Public Opinion Strategies today of 500 “persuadable” voters, finding that 64% of respondents fact-check using services like Google, 49% get their news online, and 62% trust the information they find online. Google proudly proclaimed on their blog “Internet is Key Channel To Reach Persuadable Voters,” and it was parroted in headlines around the tech and political press.

But, what does the evidence say?

First, it’s misleading to characterize the American electorate as a Utopia of open-minded, thoughtful citizens. Twelve percent of Americans still believe Obama is a Muslim (both Democrats and Republicans). The overwhelming literature in Political Psychology finds that when confronted with accurate information contradicting pre-existing false beliefs, most respondents stubbornly maintained those beliefs and, in some cases, became more convinced. “While political and religious conservative beliefs predicted a belief in Obama as Muslim, exposure to the news media did little to moderate this effect,” concluded the researchers.

Second, most respondents who self-identify as independents are actually closeted partisans. In August 2009, 82% of respondents who were independent Democrats in January 2008 continued to identify with or lean toward the Democratic Party, while 73% of independent Republicans continued to identify with or lean toward the Republican Party,” wrote Emory University professor, Alan Abramovitz.

There is evidence of a smaller but growing contingent of genuinely independent voters, with unstable party preferences between elections. Work by Russell Dalton in the recently released The Apartisan American, finds that though this faction of authentic independents does exist, they seem to be more interested in the personality profile of a candidate. Thus, even if Google is reaching this (perhaps numerically inconsequential persuadable voter), they might not be the kind of thinker who’s looking at facts objectively, anyway.

Third, surveys are a horrible, horrible way to ascertain political information. In the aptly titled, “Why Bother Asking? The Limited Value of Self-Reported Vote Intention,” Harvard Professor Todd Rogers finds that 13% of voters who say they will vote, do not vote, and 54% who predict that they wouldn’t vote, actually do vote [pdf]. Citizens are most likely terrible at reflecting on how badly their own biases influence their beliefs, and therefore would have difficulty identifying themselves as “persuadable”.

This is not to say that the Google study is completely invalid. Citizens do respond to refutation of false rumors. In an experimental focus group, two University of Florida researchers found that responses to attacks were an effective way to persuade voters, as measured by the number of participants who preferred them over their opponent after reading an attack. Of the refutation techniques studied, flat-out denials were the most effective way to respond [pdf].

Additionally, a savvy-reader may argue that presidential polls are made up of surveys, and they can be used to accurately predict elections. True, but not because voters are good at identifying their behavior, but because we have decades of matching how polls map onto actual outcomes.

We have very little evidence on what a “persuadable” voter is or their behaviors. The most likely scenario of fact-checking behavior is partisan-leaning independents cocooning themselves in an echo-chamber of their favorite blogs, which only reaffirm their existing beliefs.

So, shield your minds with knowledge and a healthy skepticism, the wielders of bad research want to own your opinions.