Algorithmic Justice

The White House released a fact sheet last week about disrupting the cycle of incarceration through their new Data-Driven Justice Initiative. There is a crisis of health in our nation’s jails, you see, and it’s costing the nation a lot of money. Which is true, but that’s not exactly the problem or solution that’s being proposed: the Data-Driven Justice Initiative aims its sights here at healthy solutions to the problem of jail overcrowding.

The three strategies

  1. Use data to identify and proactively break the cycle of incarceration.
  2. Equip law enforcement and first responders with the tools they need to respond and divert.
  3. Use data-driven, validated, pre-trial risk assessment tools to inform pre-trial release decisions.

The first of these three strategies acknowledges that the cycle of incarceration traps a lot of people who are sick. It’s based on anti-homelessness programs that target frequent ER patients, and hopes to copy their success to avoid unnecessary incarcerations.

Equipping law enforcement and first responders with tools for de-escalation is vital and should have occurred years ago. My main concern is that this solution will focus on technological innovations like apps rather than comprehensive training and making sure the training is actually used in the day-to-day.

Regarding these objective, data-driven, validated risk-assessment tools: Data-driven policy is not the same as evidence-based policy. Risk-assessment algorithms are not wiser or more objective than humans. Humans build them, feed them data, and interpret the results, and as such these tools are often subject to human biases.

Earlier this year, ProPublica put together an excellent report on the algorithms used to determine sentencing, and found in their analysis that they often were easier on white criminals than black ones. Will these risk assessment tools similarly become another source of injustice?

Similarly, evidence-based policy can be a minefield. Scientific evidence has been used unjustly to craft a number of policies that disproportionately harm marginalized groups. That being said, it can also be used to support unpopular yet functional policies such as Housing First, which is wrapped into the Permanent Supportive Housing initiative linked (incorrectly) in the White House’s Data-Driven Justice Initiative release.

I hope we move forward and use both evidence-based policy and data-driven policy tools for good. I hope that we can use data to improve the lives of people trapped in jail and those likely to be.

I worry that we are moving towards a depersonalization of criminal justice, with a focus on algorithmic efficiency rather than on empathy and human rights.

We need to keep evidence-based/data-driven policy from becoming #rationalia by centering the rights of those most affected by these policies.