Utah State University uses data to improve policing after sexual assault concerns

Utah State University’s new “Predictive Policing” model may sound like the plot of “Minority Report” at first.

But the school says it’s far less dystopian than the 2002 sci-fi thriller, and far more focused on predicting needs, e.g. For example, which football games might require more officers on patrol, or finding out where students are reporting their biggest concerns and why.

“It was phenomenal,” said Blair Barfuss, USU’s youngest police chief. “We started getting answers that I never expected. And it’s all data-driven.”

Barfuss and Erik Christensen, a senior college police officer, unveiled the model last week during a campus public safety conference that brought together administrators from colleges across the state to learn how to improve safety for students.

In recent years, the state of Utah has been the focus of criticism, particularly over how its police department and other offices on campus — as well as the Logan Police Department — have responded to sexual assaults. The U.S. Department of Justice investigated the school in northern Utah and reported in 2020 that it found “significant failures” in the school’s handling of complaints.

In December 2021, Utah grad student Kaytriauna Flint filed a lawsuit alleging the school continued to protect its football players from sexual assault claims after the DOJ report, including in her case. She pointed to footage of then-USU police chief Earl Morris and football coach Blake Anderson making derogatory comments about victims. Morris resigned, Anderson apologized, and Flint has since settled her case.

This story is one of the reasons the school brought in Barfuss to overhaul the department, although he remained in his position for a few months to attend to family medical problems. Christensen, who was previously a police investigator, has taken on the investigation of how the organization can improve.

Both Barfuss and Christensen said colleges have the resources to conduct data studies and should lead in “predictive” policing. “We have blind spots on our campus,” Christensen said. “We need to look at the numbers to address the issues.”

Here’s a look at three things USU is investigating with its new model.

1. Studying crime hotspots on campus

When classes began last fall, Barfuss said, the USU police department received a large number of calls from students and faculty who reported smelling marijuana.

“So much has been reported that we’ve had community leaders and church leaders who have reached out to our administration and want answers,” Barfuss said.

Barfuss laughed as he said that for a college campus, reports of marijuana use didn’t come too out of the blue. But things got more interesting, he said, when the department decided to map the reports.

The department expected to see hotspots on the 98 student residences on campus — the most campus residences of any university in the state.

But none of the reports came from there. The odor was reported most frequently near classroom and research buildings on the north side of campus, with some additional reports from the south end.

After a few phone calls, Barfuss and Christensen found out that USU, as an agricultural school, has a research lab that grows cannabis to study its uses. Researchers harvested the plant at the same time classes began. Then they composted it and placed it in flower beds around the north side of campus.

They also learned from the ground crew that there are skunk families living on both the north and south sides of campus, which likely led to some of the reports as well.

By sifting through the data, Christensen said, the department was able to find the true cause of the spike in reports. And they told the campus what happened — rather than just dismissing what they biased thought was the reason, he added.

The department intends to use hotspot mapping to investigate more serious issues, including assaults, in the future. If there are specific places where assaults are more common – such as the dormitories – they will try to talk to the students there about prevention and reporting.

They would also investigate if there are areas with inadequate lighting, for example, or if students are unaware that they can request a police escort. This could also help determine if it is a repeat offender, which has been a problem with Torrey Green and the fraternity assault cases.

2. How to monitor major events like soccer games – and save money at the same time

Christensen developed his own algorithm to study large campus events, starting with soccer games.

The equation takes into account the participation of participants in past games over the past six years, the number of crimes reported to the police, when a game started, and even the weather to determine the likelihood of what type of crime in a future similar game could happen.

For example, games with higher temperatures saw more medical calls for participants suffering from heat fatigue. So Christensen used the model to hire more medical officers for upcoming games where the temperature would be in the same range.

Matches against certain competitors tended to have more alcohol violations, resulting in more officers having to make arrests. The games against Brigham Young University were the lowest for that, Barfuss said with a chuckle, noting that BYU’s code of honor prohibits students from drinking. But those games, which had stronger rivalries between fans, tended to lead to more disputes.

Christensen used the algorithm to predict what might happen at last fall’s Air Force Academy game. It predicted 1.5 medical 911 calls. There were two during the game, and the staff had planned accordingly.

Data modeling also showed that more reports were called to the police towards the end of the first quarter and midway through the second quarter of a game. So Barfuss and Christensen began to adjust the number of officers they had on shifts accordingly – instead of having officers sitting around early in the game with nothing to do, which is costing the university in overtime.

It’s called surge staffing, which is also used by FedEx and UPS, Christensen said. Playing five home games last year, USU saved about $10,000 by using this model to improve games as needed.

The modeling can also be used to view events such as the Howl, the big Halloween party held annually at USU. In recent years, the event has seen a surge in reports of attacks and groping.

USU has 3,500 hours of special events each year, Christensen noted.

He can also use the data to determine the daily occupation; Most calls come in between 2pm and 6pm. This is especially helpful, Barfuss said, as police departments — especially on college campuses — struggle to afford enough officers.

Christensen found that 4.6% of the days, callers waited an average of nearly two hours for help (with priority given to more serious cases, which had shorter times). But staff rotation can address these delays. And the data can help show university administrators the need to fund more officials.

3. Planning the likelihood of certain crimes, such as drug arrests

As with soccer games, Christensen’s data analysis can trace crime rates back years into the past and predict increases or decreases in crime. 3,658 incidents were forecast for this year, down from 4,406 last year.

Christensen began investigating types of crimes. He studied drug arrests and created a model for how many USUs to expect in 2022. After ruling out runaway years like 2020, when there were fewer children on campus due to the COVID-19 pandemic, the equation predicted there would likely be 16.5 drug arrests by USU police over the past year.

By early December last year, they had 12, Christensen said, and he was concerned the analysis was wrong. He stressed that the department would not seek to get more arrests just to meet the prediction, but he retained more staff that month to meet anticipated arrest needs.

Then, at the end of the semester, four more calls came in from kids celebrating as they graduated. The department ended up with 16 drug arrests for the year, he said, “That’s a good model.”