The implicit trust the public (and the media) places in scientists and the research they conduct is never quite as shaken as when outright fraud is exposed.

Earlier this Tuesday, such an incident occurred, when it came to light that the results of a study published last December in the prestigious journal Science were likely to have been fabricated. The study in question, "When Contact Changes Minds: An Experiment on Transmission of Support for Gay Equality," was seized upon by major news outlets for its surprising and inspiring conclusions. It found that people's disposition toward gay people noticeably improved after they engaged in a 20-minute conversation with a survey taker identified as gay — more impressively, this shift in attitude lasted months later. The study seemed to validate the very best parts of our humanity. Now, its fallout might remind us of the worst.

An Unhappy Accident

David Broockman and Joshua Kalla, then graduate students from the University of California, Berkeley, unraveled the web of lies surrounding the study, authored by Michael LaCour, a UCLA graduate student, and Donald Green of Columbia University. It was an endeavor that, ironically enough, only began because they too were excited by its results and hoped to expand it further with their own pilot study earlier this May.

Yesterday they, along with Peter Aronow, an assistant professor from Yale University, published a paper laying out their case: the study, rather clumsily, had plagiarized at least part of its baseline data from a 2012 national survey conducted by the Cooperative Campaign Analysis Project (CCAP). Analyzing the shifts in attitude the LaCour study found, Broockman and Kalla found that its results were quite literally too good to be true, since it had little to none of the statistical mess that’s expected with randomized experiments or surveys. One glaring abnormality included the nearly 90 percent reinterview rate of those originally surveyed. It’s likely that LaCour lifted the CCAP data and made predictable changes to it in order to create his results, they concluded.

Perhaps most damningly of all, though, when Broockman and Kalla tried to reach out to the survey firm that supposedly helped scour for the participants in the original study, they were told that the firm had never heard of the project, nor would they have been capable of performing much of the work described in the study if they had. The two then contacted Green, who along with LaCour’s graduate advisor, confronted LaCour. Green then returned to the men on Tuesday, informing them that LaCour had admitted at least to some fudging of his numbers. He then requested that his letter to Science asking for a retraction of the LaCour study be placed in their report as well. "On Tuesday, Professor Vavreck asked Michael LaCour for the contact information of survey respondents so that their participation in the survey could be verified, but he declined to furnish this information," Green wrote.

In speaking with the blog Retraction Watch, Green elaborated on his role with the study. LaCour had come to him with the concept and some early promising findings. He then embarked on the larger study, which Green agreed to oversee. "Given that I did not have IRB approval for the study from my home institution, I took care not to analyze any primary data — the datafiles that I analyzed were the same replication datasets that Michael LaCour posted to his website," he told Retraction Watch, "Looking back, the failure to verify the original Qualtrics data was a serious mistake."

The blowback has been swift. LaCour, who wrote on his personal website that he was poised to become an assistant professor at Princeton University in July, has removed that statement. In its place is a written promise that he is gathering evidence so that he “can provide a single comprehensive response.” In the meantime, Science has issued an Expression of Concern, promising their own investigation, and many of the outlets that covered LaCour’s study with gusto have quickly placed their own updates regarding the turn of events.

As the Broockman and Kalla report note, there are still questions about other work conducted by LaCour, including unpublished data claiming to show the same effect on beliefs about abortion. Left behind in the study’s wake are the flesh and blood canvassers who actually did meet with the people LaCour had claimed he surveyed online before and after the experiment, according to interviews conducted by This American Life, which had previously run a segment on LaCour’s studies. “We had no idea Mike was fabricating data,” canvasser Steve Devine told American Life.

Perhaps the only truly inspiring part of this story is the speed with which those like Broockman, Kalla, and Green responded to the possibility of fraud. It wasn’t an oversight committee that sniffed out LaCour’s deception, but fellow graduate students. And no less than four days after the two came to Green with their concerns, he willingly asked for a retraction, unafraid of the possible repercussions to his reputation. These actions happened in the face of a wildly popular finding that made intuitive sense, and may still, as other less conclusive research shows a trend of people adopting more tolerant attitudes towards social groups they frequently come into contact with.

Disheartened as some might be by this turn of events, we should be encouraged with how it was ultimately resolved.

Source: Brockman D, Kalla J, Aronow P. Irregularities in LaCour (2014). 2015.