The level of popular acceptance of same-sex marriage has increased dramatically in recent years, but remains low in many areas. What if same-sex marriage proponents sent gay canvassers into neighborhoods to persuade opponents of gay marriage to change their potential votes on the subjects? Would this be more effective than having straight people act as the persuaders? The Los Angeles LGBT Center decided to sponsor an experiment to find out.
The experimental design was simple enough: Gay or straight canvassers were chosen randomly to visit voters in their homes and try to persuade them. The findings favored the use of gay canvassers and were so compelling that after peer review they were published in the prestigious journal, Science.
But now it appears the findings were fraudulent and the article will be retracted. The young researcher who collected the surveys measuring changes in attitudes seems to have been so determined to come up with the positive findings that he made them up, at least in part. He said he paid participants to increase the response rate, but he hadn’t. He was asked for the original data but said he had erased it. He claimed a 12% response rate, but other researchers could not achieve anything like that. He said he had worked with a survey company, but they denied any awareness of the project.
You can read more about this troubling story at http://www.nytimes.com/2015/05/26/science/maligned-study-on-gay-marriage-is-shaking-trust.html?_r=0.
1. How did this failure of honesty and openness happen?
2. Why didn’t the peer review process identify the problems before the paper was published?
3. Are you reassured that another team of researchers tried to replicate the study and ultimately found so many differences in how it was working out that they started to investigate the original study? Or are you troubled that it took this much effort to uncover an apparent fraud? Is science really self-correcting?
4. What recommendations you make to an IRB to minimize the likelihood of another such failure of oversight?