How Carceral Algorithms Distort Black Narratives: The STRONG-R Case Study

The criminal punishment system has been historically reliant on “actuarial tools” to make decisions about incarcerated people; these tools are increasingly being labeled and interpreted as “Artificial Intelligence (AI)” [1], [2]. It is ethically imperative to perform this analysis as these tools directly impact the lives and experiences of incarcerated and formerly incarcerated people. We examine how social, cognitive, and organizational processes operate in the design and implementation of these “carceral algorithms” distort what some from the digital humanities may call Black narratives [3]. In particular, we focus on a collective of such carceral algorithms called STRONG-R, that consists of pretrial, recidivism, need, classification, compliance, and low-risk screener tools [4]; the tools place their narratives in the control of the Department of Corrections. STRONG-R algorithms neglect Black narratives through quantification, and reducing human stories into numbers. We use triangulation [5], [6] to analyze the relationship between Blackness, AI, and the criminal punishment system in STRONG-R, while also situating these relationships within a history of racialized carceral systems - from slave patrols [7], to the Jim Crow laws, to the New Jim Crow [8] and the New Jim Code [9]. Ultimately, we find that the STRONG-R does in fact neglect Black narratives, and we suspect that most of these processes are occurring in the social and rational time bands. The term “Black narratives” is not meant to summarize Black experiences, but rather it is a reference to shared narratives built on history and united social experience. Celia Naylor states “I wondered how I might present a counternarrative focused on the historical roots of slavery…Daryle Williams poignantly asked: ‘How to tell a human story, in a humane way… (1) the importance of presenting stories and narratives of the lived experiences of enslaved people … (2) the necessity of the humane telling of these stories and narratives that would present the capaciousness of the lived experiences of enslaved persons and the fullness of Black humanity'” [3]. Rosa Emilia Cordero Cruz of the Cridas Project states “we've been able to uncover an example of the ways that white figures control the narratives of Black women and girls and often exercise that power by reducing them to the labor they provide that sustained their white households and socioeconomic status” [3]. The STRONG-R neglects and distorts Black narratives by ignoring the effects of anti-Blackness and ultimately cannot provide a rich account for Black experiences. Lastly, so long as the Department of Corrections is more concerned about predicting risk and security level than they are about narratives, then they will always fail to adequately represent Black people. In addition to the contextualized output from the STRONG-R process, one can also explore decision-making processes that, when combined with STRONG-R in organizational contexts, lead to these distorted narratives. An account of the decision-making processes that define the behavior of individuals and organizations using STRONG-R is also useful given the importance of individual case managers (who operate within existing Department of Corrections' contexts) to an incarcerated individuals' narrative building through quantitative translation. We are analyzing these decisions as they occur within processes on the timescale of minutes to hours, or what can be considered rational and social bands of behavior [10], [11]. As a part of thinking about behavior in the rational band, Newell [11] notes the importance of knowledge to constraining the possible paths towards goals, while also emphasizing the context-rich nature of behavior. That is, the knowledge a person brings to the situation heavily influences the path to that goal. Furthermore, “if the agent knows nothing about the outside world-its structure, history, mechanics-then it has nothing to work with in escaping from local context.” [11] What happens when case managers don't incorporate the history or mechanics of Black experience into their decision making processes and other interactions with STRONG-R? We need those involved with this translation process to be informed with more than only local, or immediate, context. Social and historical contexts should be understood and used in the decision-making processes around the STRONG-R. Considering case manager behavior along these timescales also gives the opportunity to understand the processes that lead to the problematic possibility of decisions being made at the cognitive band, which delineates “automatic” behavior in a timescale of seconds [11]. Thus, these cognitive-level decisions and characteristics regarding incarcerated individuals largely lack conscious acknowledgement of the effects of external context. This is why we must break down the implications of the timescales. In conclusion, the STRONG-R does indeed distort Black narratives, and understanding cognitive processes and timescales will help us better understand the implications of these decisions. Linda Dounia has said it well: “Black people are accustomed to being unseen. When we are seen, we are accustomed to being misrepresented. Too often, we have seen our realities ignored, distorted, or fabricated…We want the world to hold space for our humanity” [12].