Researchers at the Mandel School awarded $715,000 grant to examine thousands of police interviews with sexual assault victims
officers may use “signaling” language in sexual assault reports—occasionally
dropping hints about the validity of the victim’s claims—that possibly
influences an investigation’s outcome.
That’s according to Case Western Reserve University researchers who already studied more than 1,100 untested sexual assault kits over a 16-year period. Now, supported by a $715,000 grant from the National Institute of Justice, they will analyze more than 6,300 police reports of sexual assault in Cleveland, collected between 1993 and 2009.
The idea behind the research is to change how reports are taken to get more accurate reports and better outcomes, said principal investigator Rachel Lovell, senior research associate from the Begun Center for Violence Prevention Research and Education in the Jack, Joseph and Morton Mandel School of Applied Social Sciences.
“Responding officers can convey signaling in the narratives of police reports regarding a victim’s credibility,” Lovell said.
“For example, in general, police officers have learned that lack of eye contact during a police interview can signal deception,” she said. “But it wouldn’t be unusual for a victim of sexual assault to avoid eye contact. So that, once the ‘avoiding eye contact’ statement is included in the report, it may signal to the investigating officer that the victim is not credible, which in turn could change how the case is investigated and prosecuted.”
“We noticed very early on in our research that, at times, there would be details in police reports for sexual assault victims that seemed like the officers thought the victims were being deceptive or seemed to be confused by the victim’s behavior,” Lovell said.
Researchers will use machine-learning technology to analyze the text of the police reports to examine “signaling.” Machine-learning technology is a type of artificial intelligence that will iteratively teach computer software to detect patterns in the language used by reporting officers. These patterns will then be used in predictive modeling to forecast outcomes.
Luminais, a senior research associate at the Begun Center, said sample sizes for most
studies are usually between 200 to 300 cases.
“But this is a huge sample size, and we’re using a computer to do what would be difficult and time consuming for humans,” she said. “Our goal is to leverage our relationships and partnerships to improve how criminal justice system responds to sexual assaults.”