Most people don’t go willingly down to the basement of the Philadelphia Police Headquarters building known as the Roundhouse, home to the Police Detention Unit. They’re taken there on charges ranging from drunken driving to murder, held there for about 18 hours, and then — after a two- or three-minute hearing conducted by video conference — either released or shipped off to jail to await trial.

But Evan Dubchansky was eager to start work there.

“My idea was to intervene in someone’s circumstances as early as possible to reduce their contact with the system, knowing that the system usually causes more damage, more isolation, and more problems for the defendant than it does any good,” said Dubchansky.

Dubchansky was hired in 2018 as one of the Defender Association of Philadelphia’s first “bail advocates,” social workers who could gather information or provide referrals that might sway bail magistrates to release defendants.

It worked better than anyone expected, according to a new independent evaluation by a University of Pennsylvania researcher. It reduced new arrests and bail violations and correlated with better case outcomes, including acquittals and lighter sentences. It also accomplished a goal that has eluded other city efforts: It significantly reduced racial disparities in a system where the jail population is now 90% black or Latino.

Despite its success, the bail advocate program is no longer running. The two advocates were pulled out of the Police Detention Unit after seven months of what they described — in interviews and in formal complaints — as verbal abuse, sexual harassment, hostility, and intimidation by the police correctional officers who staff the unit.

“It was emotionally and verbally abusive, and it was what our clients experienced as well,” one of the advocates, Maggie Dekker, said, describing a degrading and frightening onslaught of comments by staff.

Dubchansky said that whenever he opened the door of his small office, which was actually a converted holding cell, “the officers would yell at me to ‘go back into my cell.’ ”

A police spokesperson said the department’s Office of Professional Responsibility initiated two investigations as a result of a complaint made in February 2019. “Both investigations resulted in findings of ‘not sustained,’ which means that there was insufficient evidence to either prove or disprove the allegations,” Staff Inspector Sekou Kinebrew said in an email.

Kinebrew said police correctional officers have since received training in the city’s sexual-harassment policies. “We are also evaluating measures to provide use-of-force training to all personnel assigned to the PDU,” he added.

He said the department welcomes the return of the program. But the Defender Association is seeking a way to do so without putting advocates back on-site, possibly by investing in a closed-circuit television system, chief defender Keir Bradford-Grey said. The agency does not currently have funding for such a system.

Paul Heaton, the senior fellow at Penn Law who evaluated the program, said it could be a preferable alternative to controversial algorithmic risk-assessment tools for a city that has committed to reducing its jail population without introducing new public safety risks.

He believes two factors powered the reduction in recidivism. The data that advocates generated helped magistrates make better, more informed decisions. And defendants’ behavior shifted as a result of that first supportive encounter in an otherwise hostile system.

“Potentially one of the things that’s happening is: If you treat people like human beings, they have a different sense of procedural justice, and they respond differently when the court asks them to do things like show up,” Heaton said.

As to how the bail advocate program reduced racial disparities, he said: “In contexts with implicit bias, one of the commonly recognized solutions is ‘Let’s slow down the process a little bit and let people think a little longer, with a richer information set.’ ”

Dubchansky and Dekker said that “slowing down” was relative. They had an average of eight minutes with each client, enough to try to calm them down, make a phone call to let a family member know where they were, learn about physical and mental health needs, and, on a good day, make a referral.

“I would hear about poverty and addiction and inequality and lack of education and housing challenges — the things that stem from the disparities in our society, that also pull people into the criminal justice system,” Dubchansky said. He’d present that information to a public defender, who would sometimes be able to argue for lower or no monetary bail.

In other cases, he’d scramble to line up a treatment bed and find a caseworker willing to pick up a defendant from custody. “It was those kinds of arrangements that had the most sway for the bail magistrates,” he said.

But at work every day, he said, he watched officers yell at and berate prisoners, punching one in the head and more kicking than nudging people slumped on the floor in opioid withdrawal. Kinebrew, the police spokesperson, said the advocates’ complaints were not sufficiently specific for the department to prove or disprove.

Both advocates said they were subjected to insults and sexual harassment. Additionally, they said, the officers made it difficult for them to do their jobs: ignoring requests to bring clients out to speak with them, preventing advocates from making phone calls to notify families of bail decisions, and in some cases, putting defendants on a transport van to prison in Northeast Philadelphia even when families were en route to post bail.

John Gross, policy director for the Defender Association, said the situation should raise alarms. “[The advocates] were not treated well, so if you extrapolate from that, how is the average person who is under arrest treated?” Gross said. “It really strips people of the idea that they can expect to be treated fairly.”

It’s one reason that Bradford-Grey, the chief defender, argues that instead of arresting every defendant, police should begin issuing summonses to those facing low-level charges that are unlikely to lead to pretrial detention.

“If we took those [low-level cases] out of the process altogether, we would have far more time to look at the middle tier who might have more options for release,” she said.

To her mind, what the bail advocate experiment underlines is both the need for early supportive interventions to help defendants stay out of further trouble and for a more comprehensive review of each case to identify those who can be safely released.

“People have been made more desperate,” she said, “just because we didn’t take the time to do a careful analysis.”