The processes for applying for, investigating, and adjudicating a security clearance in the United States are costly in terms of time and effort for both applicants and the federal government.

Interviews with applicants about their backgrounds are a key component of this process. The Performance Accountability Council asked the RAND Corporation’s National Security Research Division to develop and pilot test the potential for new interviewing methods, including machine learning (ML), to detect speech patterns that reflect deceit or truthfulness during simulated background investigation interviews. We propose that how people answer questions might be just as important as what they say during these interviews. Using this proposition, we developed and tested ML models to detect truthfulness and deceit among interviewees participating in a simulated background investigation interview.