People may lie when asked if they use pornography or drugs, but they are less apt to do so when questioned by a machine.
Using phone interviews, men and women were questioned by either a live person, a recorded human voice or a machine-like voice. The study was conducted by Mick Couper, an adjunct associate professor of sociology at the University of Michigan.
Subjects admitted to more stigmatized behaviors -- such as buying or renting pornography and smoking marijuana -- when they dealt with a recorded or synthetic voice. It didn't matter how human-like the voice was or whether it sounded male or female: The mere absence of a real person appears to prompt disclosure. Some participants known to have declared personal bankruptcy were also asked about their bankruptcy history. Nearly 20 percent denied having declared bankruptcy to a live interviewer, while less than 10 percent lied to a machine.
Twenty to 25 percent of participants hung up, regardless of whether a recorded or computer-generated voice gave the survey.
But don't expect a swarm of silicon solicitations just yet. All calls made in the study began with a live interviewer. Calls initiated and conducted by a machine may be much less effective. "Even I hang up on them all the time," Couper says, "and I am a survey researcher."