Following a resurgence of mass Black Lives Matter protests in 2020 in response to widespread displays of police violence and subsequent calls to defund police, the ACM Conference on Fairness, Accountability, and Transparency (FAccT) accepted and promoted papers about how to improve carceral technologies or make carceral algorithms “fair,” such as risk assessments for recidivism or facial recognition. Despite FAccT’s claim to care about social justice and marginalized people, these papers reinforce systems of oppression by helping the carceral state “better” police marginalized people; additionally, these papers are not authored by people in those communities disproportionately impacted by the carceral state—Black, Latinx, Indigenous, or transgender communities. This means people who are often the most distant from these issues are publishing without including the perspectives of those most impacted by them. This should not be a "hot topic" for academia: it is people's very material realities that destroy families, communities, and economic livelihoods for generations to come. Every single paper on the subject of carceral technology, no matter how they formulated their problem, aimed to reform carceral technology through "fairness." This is cruel; such an act legitimizes anti-Black, anti-Indigenous, anti-Latinx, transphobic, queerphobic, and ableist carceral state violence.

The venue claims to hold space for "participatory research" with people on the margins of society while simultaneously never asking them whether or not they want carceral technologies to be reformed. And when we say we do not want them reformed, nobody listens to us. Far from being neutral or “more objective” than police, these technologies reinforce systems of oppression by helping the carceral state “better” police marginalized people. However, this should not be seen as a cause to "improve" carceral technology, but rather more reason to remove these technologies completely from our communities.  Instead, we should focus our work on highlighting the injustice of carceral technology. In this way, money that would have gone towards funding carceral technologies could instead be devoted towards education, economic opportunities, and social and community programs.

Call to Action: I call on ACM FAccT to create a policy rejecting papers that reinforce carceral violence against marginalized people, including carceral risk assessments (e.g., recidivism), facial recognition technology, predictive policing, etc. In addition, it should reject any submissions testing their methods on carceral data, such as the COMPAS recidivism risk dataset. Reforming carceral technologies is not transformative for those who are oppressed. Lastly, given FAccT’s previous involvement in perpetuating carceral technologies, it should make an effort to donate to Black, Indigenous, Latinx, and trans abolitionist organizations.

Carceral Technology Papers at FAccT 2021:

Fairness in Risk Assessment Instruments: Post-Processing to Achieve Counterfactual Equalized Odds

Fair Classification with Group-Dependent Label Noise

Towards Fair Deep Anomaly Detection

Bridging Machine Learning and Mechanism Design towards Algorithmic Fairness

The effect of differential victim crime reporting on predictive policing systems

A Bayesian Model of Cash Bail Decisions

Avoiding Disparity Amplification under Different Worldviews

A Statistical Test for Probabilistic Fairness

Leave-one-out Unfairness

Value Cards: An Educational Toolkit for Teaching Social Impacts of Machine Learning through Deliberation

Detecting discriminatory risk through data annotation based on Bayesian inferences