Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Software bug release dates keeping hundreds, In 2019, the state’s laws had changed, but the state’s software suite hadn’t.
A report from a “whistleblower” says that important prison management software was not kept up to date, so hundreds of inmates in Arizona were locked up longer than they should have been.
An NPR station in Phoenix, KJZZ, said that employees of the Arizona Department of Corrections have known about the bug since 2019. The software can’t handle a change in state law because it hasn’t been updated. This is what caused the bug.
Drug possession convictions are a big reason why Arizona has one of the highest rates of people in jail or prison in the country. In June 2019, a change (PDF) to Arizona state law made it possible for prisoners convicted of certain nonviolent drug crimes to get credits toward early release. Inmates who are eligible and finish a program while they are in jail, like a GED equivalent or treatment for drug abuse, can get three days of credit for every seven days they have already served. This means that they can only go to jail for 70% of their sentence.
The Arizona Department of Corrections says in the program’s own Frequently Asked Questions (PDF) that neither inmates nor their families should contact anyone to ask for a review of their eligibility. “This is done automatically by system programming, which makes a list for Time Comp,” the document says. “The reviews are being written based on when they are supposed to come out.” On the other hand, the whistleblowers say that the system can’t find eligible participants and can’t do the math.
“We knew from the start that this wasn’t going to work,” a source from the Department of Corrections told KJZZ. When they passed that bill, we looked at it and said, “Oh, shit.”
People who came forward said that they started giving “repeated internal warnings” to the IT officials of the department in 2019. A bug report from October 2020 was sent to KJZZ. It said that the ACIS software doesn’t keep up with changes in the law.
“This calculation is not in ACIS at all right now,” the report says. “ACIS can figure out that you get 1 credit for every 6 days you serve, but this is a new way to do it.”
KJZZ was told that the problem with ACIS is known to the department. A department representative, Bill Lamoreaux, told KJZZ that data is being calculated by hand and then put into the system. At least 733 prisoners are eligible for the early release program, but they haven’t signed up yet.
Software to manage (or not manage) jails
In the US, “corrections” software is a big business. On one level, jail and prisoner management systems make sense: you need to know who is in a facility, where they are, what their medical needs are, who is representing them in court, etc.
As more and more parts of the criminal justice system are given to algorithms to run, more and more real people are being hurt by bugs in these software packages. Local news outlets at the time said that putting ACIS in place in Arizona took three years longer than planned and always cost more than planned.
Sources have told KJZZ that ACIS has been full of bugs ever since it was finally put to use over the Thanksgiving holiday in 2019. Several parts of the software haven’t worked right, including parts that keep track of inmates’ health care, head counts, property, commissary and financial accounts, religious affiliation, security classification, and gang affiliation. And if a mistake is made in a prisoner’s file because of human error, it may not be possible to get rid of or fix the mistake.
“In one case, a prisoner’s record was marked with a disciplinary action by mistake,” a source told KJZZ. “You can’t get it back, though. So, that guy was punished and couldn’t call anyone for 30 days. These are the kinds of things that bother you every day.”
In the past few years, many people have said bad things about the software suites that police departments and courts use. People use these suites long before they have to deal with software for running a prison. In 2016, ProPublica released a long, detailed report that said the courts’ risk assessment software said that Black suspects accused of similar crimes were more likely to commit more crimes than white suspects accused of the same crimes.
In 2018, researchers looked at the software and found that groups of random amateurs could predict how often suspects would commit crimes again just as well as the software. About a third of the time, both people and software were wrong.
READ MORE ARTICLES:
