CSU professor uses AI to predict outcomes in rape cases – The Cauldron

Dr. Rachel Lovell, director of the Criminology Research Center and assistant criminology professor at Cleveland State, was recently published in the Journal of Criminal Justice with a pair of articles sampling 5,638 rape kits and cases within the Cleveland Division of Police from 1992 through 2012.

Both articles use artificial intelligence to assess the same sampling of rape cases with the Cleveland Police, specifically looking at victim credibility. Lovell did not accomplish the feat alone. 

“I want to highlight the fact that I have an amazing team of researchers who did the majority of the day-to-day type of work, many of whom had to read rape report after rape report after rape report, and that certainly takes a toll on emotional and physical health,” Lovell shared. “It was a big team, a collaborative effort.” 

Lovell’s team included some individuals with backgrounds in criminology, sociology, sexual assault and rape reports. Other team members were computer scientists familiar with artificial intelligence methods.

“It was a great combination of our skill sets,” said Lovell. 

A CSU professor’s research on rape cases form the Cleveland Division of police suggests reforms in police report writing. | Photo Credit: Tim Evanson, Wikimedia Commons

Beginning of Project

The research was a natural expansion of the Sexual Assault Kit Initiative, which members of the team were already working on. After recognizing an innuendo, or a signaling about a victim’s credibility, in police reports on rape while reading through each report individually, the team wanted to do the work on a massive scale. 

Without artificial intelligence, “it would have been insurmountable to do,” said Lovell.

Lovell also explained that human brains do not work in the same way as computers and can miss connections. For example, humans may have overlooked phrases such as “woman was raped” or “ORC 2907.02 Rape” or other variations of those words. The team’s programmed artificial intelligence recognized that police reports using these phrases had a more successful case outcome. 

Reports that were longer and included more of the victim’s perspective written into the report were more likely to be recommended for prosecution. 

“The more negative the report, the more likely it was to move forward. But the negative part was not about the victim, it was about the nature of the sexual assault,” added Lovell.

Next Steps

Through connections with the Sexual Assault Kit Initiative, the team is sharing the findings with training agencies for law enforcement officers, including the International Association of the Chiefs of Police. 

Lovell explained that police officer training can be improved to better serve victims, especially those of rape. Officers are currently trained to write all reports the same way, no matter the context of crime. 

“Crimes of interpersonal violence and violence should not be written in the same way as property crime,” suggests Lovell.

Details are essential, according to Lovell. When looking at rape reports, the elements and background of the crime must be included so the prosecutor knows how to charge it, but details on penetration, force, threat of force, or incapacitation were missing from many reports in the study. Saying “he had sex with” is not the same as “he forced.” 

In addition to restructuring and improving training, the technology could also be utilized to assist police officers. Lovell explained that officers currently have a sort of “open text box” without prompts, structure or guidance when writing reports. A prompt to intentionally include information on the elements of the crime, when that information is available, could standardize police report writing to make it more efficient and reduce bias.

“Reducing the bias in our writing can also help us reduce bias in our interactions and our thought processes,” explained Lovell. “It also has potential to make better interactions with victims because the way we think influences the way we write, and the way we write also influences the way we think.”

At the time the project began, ChatGPT did not exist. Lovell’s next steps include researching how to automate some of the writing processes for police reports, while still maintaining the privacy of the criminal justice information.

Future Research

With these findings, there are different directions future researchers may take. Black teenage girls were the largest group of victims in the reports, which Lovell plans to explore in more detail.

Teenage girls also make up a large portion of runaway cases, which the team may also explore in future research. 

A final direction stemming from this study is unfounded cases, where police determine that a crime did not occur after investigation. 

“We can read the reports and in a good portion of those, there is really no evidence to suggest that a crime did not occur. In many of them, it is that the victim did not know what happened to her, but that does not mean it did not occur,” said Lovell. 

Artificial intelligence could be adapted to each of these directions, combining qualitative data, quantitative data and machine learning technology for multi-level analysis. 

Lovell’s Background

While Lovell is not a survivor of sexual violence, she has yet to meet a woman who does not have some sort of a “Me Too” experience. 

“I certainly draw inspiration from the survivors to say that for many of them, no one spoke up for them. These are survivors whose kits were shelved, whose reports were closed very quickly, and whose cases were closed very quickly. No one listened to them,” said Lovell. 

She has been working in gender and violence since 2009. Lovell began her work with the Sexual Assault Kit Initiative in 2015, finding the movement an inspiration. 

“Much of the [Sexual Assault Kit] Initiative is led by survivors, and you can really see this galvanizing force of the survivors and the power they have, gaining empowerment from what happened to them and being a force for change,” added Lovell.

Though this project has ended, Lovell mentioned opportunities for students to get involved in future projects through the Criminology Research Center at CSU.