Evidence of What, for Whom? The Socially Contested Role of Algorithmic Bias in a Predictive Policing Tool

Algorithmic bias isn't a neutral technical problem: Different stakeholders in Chicago—community organizations, researchers, and police—interpret the same evidence of bias in the predictive policing algorithm in fundamentally different ways, using it to advance their own distinct goals around criminal justice reform or police authority.

"Evidence" serves political purposes: The paper found that stakeholders strategically deploy findings about algorithmic bias as evidence to support pre-existing positions, whether that's reforming police patrol policies, rejecting algorithmic tools entirely, or reaffirming existing power structures—meaning the same technical flaw can be weaponized for opposing agendas.

Community-centered uses of bias evidence focus on healing and structural change: Community organizations used algorithmic bias findings to reframe crime as a structural problem requiring community repair and healing, while more powerful actors (like police departments) used the same evidence to maintain or justify existing systems of authority.

Technical fixes alone won't resolve the underlying conflict: The paper reveals that disagreements about predictive policing tools reflect "long-standing tensions" between different values in criminal justice reform—particularly between liberation-focused and healing-focused approaches—that can't be solved by simply removing algorithmic bias from the tool.

Who gets to define "the problem" matters enormously: The study demonstrates that without explicit attention to whose values and definitions are centered, algorithmic bias can become a tool that actually reinforces existing power imbalances rather than challenge them.