Is predictive policing software making the justice system more biased, not less?

Lens adapted to this topic: Potential benefits and reform

A serious alternative view is that the central problem is not algorithms as such but how agencies select data, define risk, deploy predictions, and oversee decisions. On this account, properly designed tools might make resource allocation more consistent or reveal patterns missed by unaided judgment. Some critics of categorical rejection argue that fairness constraints need not necessarily reduce predictive accuracy. This view nevertheless accepts that current evidence and governance are inadequate.

• Proponents argue predictive systems may forecast crime more accurately or allocate police resources more efficiently than traditional methods.
• Algorithmic tools can be assessed and reformed rather than rejected outright; stakeholder responses include changing patrol policy and using bias evidence to improve interventions.
• A critique of the dominant fairness debate argues that fairness constraints do not necessarily require a loss of predictive accuracy in criminal-justice models.
• The alternative case remains conditional because available effectiveness research is sparse and the evidentiary record is contested.
• Even this reform-oriented position must address what is being predicted, whose data define risk, and how decisions are independently reviewed.

6 sourcesView sources