Law as Ideal Types

Ideal-typical analyses tend to accentuate law’s internal coherence at any point in time. Through such strategic simplification or “one-sided accentuation,” this perspective produces “a unified analytical construct” (Coser Reference Coser2003, 223) that renders complex realities legible for identifying how “legal modes” vary in relation to phenomena such as the economy, religion, or political power (for example, Weber Reference Weber, Rheinstein and Edward1954; Nonet and Selznick Reference Nonet and Philip1978; Kagan Reference Kagan2001). Ideal-typical perspectives have also defined orienting works in punishment and society research from its inception (Durkheim Reference Durkheim1973; Rusche and Kirchheimer Reference Rusche and Otto2004) to the present (Beckett Reference Beckett1997; Simon Reference Simon2007; Wacquant Reference Wacquant2009; Alexander Reference Alexander2010). In general terms, such accounts trace large-scale shifts in “modes” of penal power and their ideological underpinnings. Thus, Malcolm Feeley and Jonathan Simon (Reference Feeley and Simon1992) have characterized a “new penology” in contrast with the old, and David Garland (Reference Garland2001) has explored a shift from the “penal-welfarist” model to a “culture of control.” Ideal-typical perspectives similarly portray the Sentencing Guidelines as a critical intervention point that initiated a “revolutionary new system for punishment of federal crimes” (Stith and Cabarnes Reference Stith and Cabarnes1998, xi; see also Bowman Reference Bowman1996; Espeland and Vannebo Reference Espeland and Vannebo2007). Such explanations cast the 1980s sentencing reforms in terms of a paradigm shift from an “indeterminate” to a “determinate model” (Nagel Reference Nagel1990; Reitz Reference Reitz, Petersilia and Kevin2012), part of a “strategic turn” toward managing aggregates (Feeley and Simon Reference Feeley and Simon1992, 461) or the imposition of a rigid, “neoclassical logic” on a field infused with substantive justice concerns (Savelsberg Reference Savelsberg1992). To be sure, ideal-typical perspectives do not necessarily ignore local variation. Instead, they situate it within “a broader picture” that, by and large, assumes a coherent form (Feeley and Simon Reference Feeley and Simon1992, 449, n1).

Law as a Field

Over the past four decades, socio-legal work increasingly has employed the metaphor of fields to make sense of law and legal action. Drawing from various traditions—notably, organizational theory and Pierre Bourdieu’s work—field perspectives typically critique ideal-typical accounts for being more descriptive than explanatory. As a corrective, field perspectives focus inward and downward: away from the macro/global and toward the meso/local (R. Scott Reference Scott2013; compare Fligstein and McAdam Reference Fligstein and Doug2012). As John Martin notes, “a field replaces the idea of action at a distance, in which X somehow directly affects some Y that it does not touch, with a purely local explanation” (Reference Martin2003, 7; emphasis in the original). Demonstrating the “loose coupling” between institutional design and implementation (Meyer and Rowan Reference Meyer and Brian1977), field perspectives focus on field-level rules, norms, and regularities: the semi-autonomous vernaculars that translate large-scale shifts into actual outcomes. Like in magnetic fields, exogenous vectors get warped by a field’s endogenous forces upon entering its perimeter (Bourdieu and Wacquant Reference Bourdieu and Loïc1992, 17). The methodological orientation of field studies therefore leans toward smaller scales and shorter time frames.

This emphasis on intra-field “translation” entails several implications. First, field studies typically conceive of change as occurring in analytically distinct stages: an external change occurs; internal field processes translate; local outcomes follow (Garland Reference Garland2013, 483). In this vein, Philip Goodman, Joshua Page, and Michelle Phelps (Reference Goodman, Joshua and Michelle2017, 13) offer the metaphor of “tectonic plates” to explain how macro-level “trends” shift under the field and present opportunities on which field actors variably act. According to field perspectives, if “extra-field” processes are to affect legal action they have to “work through” (Heimer Reference Heimer1999), be “refracted by” (Garland Reference Garland2001), become “endogenized” (Edelman Reference Edelman2016), or be “translated” into (Page Reference Page, Simon and Sparks2013) locally established practices. Second, the focus on local translation opens space to consider how the same “exogenous” phenomenon (for example, new legislation) variably manifests in different locales (Merry Reference Merry1988; Morrill and Musheno Reference Morrill and Michael2018). This view therefore critiques the “pyramid” metaphor of for dispute resolution and adjudication (Miller and Sarat Reference Miller and Sarat1980) for ignoring “the potential for legal pluralism in which disputes proceed simultaneously along many coexisting paths with multiple normative systems for resolving conflict” (Albiston, Edelman, and Milligan Reference Albiston, Edelman and Joy2014, 106). What once appeared as a centralized-hierarchical structure, with a judge at the top, transmogrifies into a “dispute tree” nestled within a decentralized “legal forest.”

Field perspectives have increasingly gained conceptual currency in punishment and society research (Sudnow Reference Sudnow1965; Blumberg Reference Blumberg1967; Feeley Reference Feeley1979; Hagan, Hewitt, and Alwin Reference Hagan, Hewitt and Alwin1979; Kohler-Hausmann Reference Kohler-Hausmann2018; Rubin Reference Rubin2021), and much literature has similarly explored local variance under the homogenizing regime of the Sentencing Guidelines. This work traces intra-field variation to “localized” understandings of “going rates” (Ulmer Reference Ulmer1997), “court communities” with “localized legal cultures” (Eisenstein and Jacob Reference Eisenstein and Herbert1977; Church Reference Church1985; Dixon Reference Dixon1995; Kautt Reference Kautt2002), courts as “organizational domains” (Johnson, Ulmer, and Kramer Reference Johnson, Ulmer and Kramer2008; Ulmer and Johnson Reference Ulmer and Johnson2017), or the Guidelines’ incorporation into “local narrative forms” (Lynch Reference Lynch2019). Thus, Mona Lynch and Marisa Omori (Reference Lynch and Marisa2014) demonstrate how federal sentencing practice not only diverged across place but also became variably entrenched over time such that top-down reform was reappropriated in local practices. Revealingly, they “view the federal district court system not as a singular national legal structure with hierarchically arranged and geographically dispersed subunits” (pyramid-like system), “but rather as a semi-autonomous set of systems governed by the same formal rules, statutes, and procedural policies, while also embedded in localized legal cultures that are themselves shaped by regionally specific historical contingencies and norms” (412; emphasis added).

Field Assembly

Field assembly involves the process of specifying the data format and creating a field panorama in which practices can be located. This foundational process entails material decisions that derive from a certain imaginary. Since these decisions set datafication in motion, the ensuing process will encode the underlying imaginary into all future data, whose supposed neutrality will then bestow objectivity on future analyses such that the emergent data and field panorama appear given and universal. The ideological underpinnings of datafication are particularly visible during this foundational stage. Once field assembly is complete, the passage of time and accumulated investment in already-existing formats make the contrived aspects of datafication less visible and harder to alter (Thévenot Reference Thévenot1984). As Wendy Espeland and Michael Sauder (Reference Espeland and Sauder2007, 19, n21) note, “commensuration is most jarring and seems less like measurement and more like artifice when it encompasses things to do not already seem alike.” Tracing datafication’s early stages helps foreground the context and people involved in synthesizing a digital ontology by which different things will become alike. Accordingly, I now turn to the work of the commission between October 29, 1985, and April 13, 1987 “to create a set of guidelines” (Breyer Reference Breyer1988, 6). I pay particular attention to how each stage assumes inheritances from antecedent action and anticipates subsequent action (Introna Reference Introna2016), emphasizing the links that extend from contingent historical circumstances, to a certain imaginary, to specific instruments, to the assembly of a particular panorama.

The co-productive relationship between the Sentencing Guidelines and carcerality began with a foundational decision, described by Justice Stephen Breyer (Reference Breyer1988, 18)—then a Circuit judge and member of the first commission—as the “past practice compromise.” In essence, since the commission failed to reach a normative agreement about the Guidelines’ “philosophical premises” (15), it decided to side-step moral and political dispute by resorting to simplified empirical analysis, indeed to “base the Guidelines upon typical, or average, actual past practice” (17). In the commission’s words, it “sought to resolve the practical problems of developing a coherent sentencing system by taking an empirical approach that starts from existing sentences” (USSC 1987b, 16; emphasis added). Instead of positively articulating a vision for a “better future,” the commission thus decided to tether the future onto the past by “replicating the average pre-Guidelines sentences handed down to particular categories of criminals” (Breyer Reference Breyer1988, 17; emphasis added). The goal was to deploy this “empirical approach” to identify aggregate patterns and “limit the otherwise broad range of sentences that may be (and currently are) imposed” (USSC 1987b, 17). Far from being a mere technical compromise, this procedural escape hatch from normativity drew legitimacy from an increasingly pervasive imaginary that views technology as an instrument for maintaining homeostatic time instead of for advancing progressive time. This approach embodied “an internal regulation of governmental rationality” (Foucault Reference Foucault and Burchell2008, 10), one arrived at through studying the target population and “letting the data speak for itself” (Mayer-Schönberger and Cukier Reference Mayer-Schönberger and Kenneth2013, 6–12). The Guidelines were thus premised on the ambition to “establish a sort of homeostasis, and compensate for variations within this general population and its aleatory field” (Foucault Reference Foucault and Macey2003, 246). This signaled a foundational shift—not searching for the “why” but only the “what” or shifting from justification to pattern recognition.

Importantly, it was no historical accident that the commission decided to base its guidelines on this homeostatic imaginary. Given a pervasive racialized “tough-on-crime” agenda, on the one hand, and the commission’s mandate to redress “unwarranted sentencing disparities,” on the other hand, the “past practice compromise” offered a colorblind harbor (Murakawa and Beckett Reference Murakawa and Katherine2010). This was so given the immanent duality of homeostasis: whereas, for some, it promised stability and safety, for others, it meant continued oppression that could now be justified by the simple fact that it had become the norm. The commission’s logic could thus ensure the colorblind enactment of a “kind of temporal penitentiary in which oppressed people are locked into a dystopic present” (Benjamin Reference Benjamin2016, 150). This imaginary was politically effective precisely because it involved this duality by which neutrality (in the sense of deferring normative accountability) becomes indeed carceral. Thus, typical past practice became a normative compass instead of that which needs to be thwarted.

Having reached this decision, the commission now turned to creating a lens by which to see the past. How, then, did the commission operationalize its “empirical approach”? How precisely did it transform “existing sentences” into data? As noted, the starting point for determining “offense levels” and corresponding “ranges” emerged from a statistical analysis of “recent sentencing practices” (USSC 1987b, 21–26). This meant, first, amassing information about past practices and, second, contriving methods by which to make this information into analyzable data. As for the first task, the commission accumulated information from several different sources. First, the Administrative Office of the US Courts provided the commission with a “computer file of all defendant records in [the] Federal Probation Sentencing and Supervision Information System” (21). This database included almost one hundred thousand felony and serious misdemeanor cases leading to convictions since mid-1983. Since this database was incomplete, the commission collected an additional sample of 10,500 case histories about which it received further information and complete presentence investigation reports from the Probation Division and the Bureau of Prisons. The commission thus gained “ready access to quantitative and qualitative information in the form of 10,500 computer records and even more detailed information in the form of 10,500 presentence reports” (21). In so doing, it gathered information that was originally created by different institutions for a variety of reasons, for the purpose of identifying patterns within the target population.

The next step was to develop a method by which to tabulate or remake divergent objects (“records,” “case histories,” “presentence reports”) into data and, as it were, to let the data speak. As Justice Breyer (Reference Breyer1988, 7) has summarized, the information that the commission gathered was “entered into the Commission’s computers and used to determine what factors typically account for more, or for less, time actually served.” To enter divergent objects into its computers, however, the commission first needed to make consequential decisions, such as specifying what factors to include in the data form and, importantly, which to exclude from the frame. Though the precise coding scheme is not delineated in the report, the commission recounted that its central goal was to identify “typical” sentences by “baseline offenses” and certain “aggravating and/or mitigating factors.” To this end, “the Commission relied on standard statistical techniques (multivariate maximum likelihood estimation) to infer how [first-time] defendants typically would be treated given prevailing sentencing and parole tactics” (USSC 1987b, 23; emphasis added). Conceding that “no statistical model could replicate judicial decision making,” the commission nonetheless stated that such a model could provide “a meaningful synopsis of current sentencing practices” (emphasis added). The results were summarized in a form known as the “Levels Tables.” Figure 2 summarizes the estimated time served for baseline offenses, while Figure 3 summarizes estimated levels adjustments by offense circumstances.

FIGURE 2.

Estimated Time Served for Baseline Offenses: 1st Time Offenders, Convicted at Trial, Sentenced to Prison, Adjusted for Good Time.

Credit: USSC 1987b, 27.

FIGURE 3.

Estimated Level Adjustments: 1st Time Offenders, Convicted at Trial, Sentenced to Prison, Adjusted for Good Time.

Credit: USSC 1987b, 36.

To provide a “synopsis of current sentencing practices,” then, trained officials collected information from a variety of sources; read through it with a targeted eye, coding for certain features only and ignoring the rest; tabulated the output into one database; and analyzed the emergent data so as to trace aggregate patterns. This undertaking involved “a system for discarding information and organizing what remains into new forms” (Espeland and Stevens Reference Espeland and Stevens1998, 317; Espeland and Vannebo Reference Espeland and Vannebo2007, 28). A crucial point in this respect is that the format into which past practice was translated was not itself part of empirical reality; rather, it derived from a colorblind-carceral imaginary, itself fashioned under contingent historical circumstances. Accordingly, it seems more adequate to speak of the emergent data not as “given” but, rather, as “achievements” (Latour Reference Latour1999, 42). In this way, a particular ideology prefigured the assembly of a technical instrument for translating past practices, which had much to do with identity and context, into a digital syntax that sees only baseline offenses and certain factors. This instrument was effective in creating a colorblind-carceral panorama. If under indeterminate sentencing factors such as race, socioeconomic status, disability, or pre-conviction information influenced officials (American Friends Service Committee 1971), then the commission’s “empirical approach” reconfigured discriminatory practices through a colorblind lens. In deciding, for instance, to make “baseline offense” an independent variable, without disaggregating it from factors such as race or socioeconomic status, the commission de facto prioritized consistency (which includes maintaining past discriminatory patterns) over equity (redressing them). Datafication thus absorbed legacies of oppression and effectively reconfigured them into a colorblind scheme (Barocas and Selbst Reference Barocas and Selbst2016).

Once a panorama of the past had been assembled, the commission could utilize it for constructing a view of the present. In this sense, the commensuration of past decisions proceeded with an eye toward classifying present decisions as legitimate or illegitimate. Already in its 1988 report, the commission relied on a random sample to analyze the frequency of judicial “departures” from the Guidelines, defined as cases wherein judges sentenced outside the calculated range, regardless of the reasons given. The commission reported that, of the 2,258 cases in a random sample, 82.3 percent of sentences were “compliant” in that they fell within the applicable range, 2.9 percent were above the range, 5.7 percent were below the range because the defendants assisted the government, and 9.1 percent were below the range for other reasons (Tonry Reference Tonry1989). In this way, commensuration simultaneously made sentencing decisions alike, by translating them into a standardized format, and distinguishable, in that they could take on such qualities as “compliant” or “noncompliant” (Freed Reference Freed1992, 1720–21). From then on, every decision could fit into one of two categories, and this classification would enable numerous subclassifications, analyses, and interventions.

If commensuration was done with an eye toward classification, classification was done with an eye toward regulation. Thus, the SRA expressly charged the commission with a duty to periodically “review and revise [the Sentencing Guidelines] in consideration of comments and data coming to its attention.”Footnote ⁶ As Breyer (Reference Breyer1988) attested, “the system is ‘evolutionary’—the commission issues Guidelines, gathers data from actual practice, analyzes the data, and revises the Guidelines over time.” Initially, this “evolution” proceeded through formal law: the commission issued formal amendments that designated off-limits factors, which the courts had declared, through judicial interpretation, as valid grounds for departures (M. Miller Reference Miller1992, 445, n132). This traditional dynamic of overt contestation over the formal meaning of law by two sovereign entities still occurred at the level of justification. The commission’s work, however, soon devolved into the less conspicuous level of pattern recognition. This shift was exemplified in the commission’s 1993 establishment of a working group tasked with studying departures from the Guidelines to “determine whether current departure practice is consistent with statutory authority and congressional intent” (USSC 1994, 8). For this purpose, the group undertook “an empirical analysis of district court departure practices, reviewing in detail 1,400 departure case files” in an effort to “promote better understanding of the role of departures in Guidelines sentencing.” In this way, the data format that the commission created for rendering “past practice” visible, which itself derived from a carceral-colorblind imaginary, was now utilized to view and regulate the present. In this vision, judges’ interpretations and precedents mattered less than aggregate patterns, and judicial practices were made into regulatable phenomena. In effect, this vision meant fastening the future of sentencing onto its past in ways that absorbed historical inequities, legitimated their cumulative effect, and assembled an apparatus that no longer depended on present human prejudice for enacting disparate impacts (Benjamin Reference Benjamin2019a).

Having created a view of sentencing decisions in terms of classifiable, regulatable objects, the commission could gradually assemble a more intricate and gapless panorama. As early as 1990, the commission began to disaggregate judicial departure rates by “jurisdiction,” “reasons for departure,” and “primary offense category.” With time and technological advancement, these operations grew more elaborate and comprehensive. In 1990, the commission still relied on a statistical sample (25 percent) from a pool of only ten districts to produce departure estimates. As Figure 4 demonstrates, this sample-based analysis provided only a partial representation of the field, not yet a holistic panorama. A year later, however, “submission rates” were sufficiently high (92.4 percent) to report departure rates not based on samples but, rather, for all cases received (USSC 1991, 133). These developments, conjoined with the subsequent employment of geo-visualization, signaled a gradual shift toward big data logics (n = all) that allowed the commission to assemble a gapless panorama (Figure 5).

FIGURE 4.

GUIDELINE DEPARTURE RATES FOR SELECTED DISTRICTS’ (October 1, 1989 through September 30, 1990).

Credit: USSC 1990, 81.

FIGURE 5.

Downward Departure Rates by Judicial District, Fiscal Years 1991 versus 2001.

Credit: USSC 2003, 33.

Here again, a concern for governing undergirded classification. The commission demarcated judicial subpopulations as objects of analysis and relied on the knowledge thus produced to govern practice in varying degrees of aggregation. The maps in Figure 5 demonstrate this point. In geo-visualizing district-level increases in departure rates over time, the commission datafied growing trends and made them legible and actionable for long-distance governing. The sub-setting of departure rates by district gradually institutionalized “districts” as units of comparative analysis, and thereby justified variable degrees of analytic attention and regulative interventions. Furthermore, the commission soon employed yet another layer of classification, sorting districts into two distinct categories of either “high” or “low” departure rates. Judicial districts were effectively assigned a “risk-level” based on aggregate compliance rates. Not unlike longitudinal studies of offenders’ risk levels (Wolfgang, Figlio, and Sellin Reference Wolfgang, Figlio and Thorsten1972; Greenwood and Abrahamse Reference Greenwood and Abrahamse1982), this classification proved stable over time as “districts classified as having relatively high downward departure rates, and districts having relatively low downward departure rates, generally remain in those categories from year to year” (USSC 2003, 34).

My intention in detailing these operations is not to “reveal” unknown facts or impugn the commission’s integrity. Rather, I aim to foreground “backdoor procedures” through which ideology prefigures datafication. With the resulting apparatus in place, commissioners and judges need not harbor racist or classist sentiments (whether explicit or implicit) to continue to enact past inequities. They simply need to do their job. As Benjamin (Reference Benjamin2019a) notes, “[t]he power of the New Jim Code is that it allows racist habits and logics to enter through the backdoor of tech design, in which the humans who create the algorithms are hidden from view.” Nevertheless, the process of field assembly still targets primarily the level of representation—namely, of assembling a coherent image out of disparate practices. This mechanism does not explain how datafication influences on-the-ground practice or enacts a data environment into being. For a field panorama to “make a difference” in practice, it must become effectively linked to its object. Datafication extends such performative links through the mechanisms of symbolic projection and material inscription.

Symbolic Projection

Symbolic projection involves the process of mediating a field panorama from one site to removed actors’ ways of seeing. This mechanism works by affecting cognition, changing the locus and form of attention, and creating new connections while obscuring others (Espeland and Sauder Reference Espeland and Sauder2007, 16). By producing a “synopsis of sentencing practices,” the commission has also rearranged the meaning of sentencing around the syntax by which this very synopsis was fashioned. Creating the Guidelines therefore meant not only their imposition as formal law but also the rearticulation of the task of sentencing in ways that (1) diverted attention from vast amounts of information and (2) arranged what remains within a standardized form. While this new form was immediately authoritative in the legal sense, its cognitive authority accrued with time as it rendered less visible and more “unreal” anything that was hard to express in its language (Simmel Reference Simmel, Bottomore and Frisby1978, 443–45).

Symbolic projection takes effect through various media, including objects such as manuals and data collection forms (more on this below), language, the circulation of data analyses, and people. A good example is the classification of judicial decisions in terms of “compliant” or “non-compliant.” The outrage evoked by this nomenclature attests to its symbolic efficacy (Weinstein Reference Weinstein1992). Yet even after this evocative language was changed, the definition of “departure” remained contested. As Daniel Freed (Reference Freed1992, 1721; emphasis added) notes, “[t]he distinction between compliance and departure was a false dichotomy, manufactured by the Commission to deter judges from sentencing outside the prescribed ranges. It understated the considerable opportunities for discretion available to participants in the sentencing process.” While the “manufactured” nature of this distinction was more conspicuous at first, the notion of “departures” as defined by the commission gradually settled and became an obvious part of sentencing (Berman Reference Berman2000). Another example of symbolic projection is training sessions that instructed new judges on the administration of the Guidelines. As two instructors reported, these sessions affected new judges such that they were often more concerned by how to manage the little discretion the Guidelines left them than by considerations of fairness and justice (Mihm and Gertner Reference Mihm and Gertner1998).

Another important medium of symbolic projection involves data analyses and visualizations. Figures 4 and 5 provide helpful examples. While, on one level, these figures simply describe practice from a bird’s-eye view, on another level, they may affect judicial perceptions. A judge who makes sense of these figures must concede, however temporarily, the ontology of its composite elements: sentencing decisions as binary variables, judges as data points, and judicial populations as entities that can take on such values as high or low departure rates. And given that “seeing bestows on objects an accent of truth,” these visualizations confer upon their classificatory operations a sense of objective reality (Amann and Knorr Cetina Reference Amann and Knorr Cetina1988, 134). To interpret these data, in other words, means seeing like the commission. With time and repeated exposure, judges were able to locate themselves within a growing data environment and become increasingly concerned with converging toward the mean.

Material Inscription

Material inscription is the process by which a field panorama palpably inserts itself into the very field it assembles through delegating action onto objects and rearranging the internal composition and external environment of practice. The USSC’s (1987a) Federal Guidelines Manual provides a key example. If, in the process of fashioning the Guidelines, human actors laboriously reformatted past decisions into data, the resulting manual seamlessly crystalized the entirety of this labor in a material object with the capacity to order future decisions. Thus, an innocent table freezes an entire imaginary and the social relations from which it derives (Figure 6). Any future utilization of this immutable object will mobilize all the labor and translations inscribed therein, regardless of the individual judge’s subjective orientations (Latour Reference Latour1999).

FIGURE 6.

1987 Federal Guidelines Manual.

Credit: USSC 1987a.

Unlike symbolic projection, material inscription does not require actors to alter their deliberative frames (Espeland and Stevens Reference Espeland and Stevens1998, 322). All they need to do is follow routinized procedures. The Federal Guidelines Manual holds an undeniable material presence: the first manual was 329 pages long and comprised seven chapters and two appendices. More recent manuals are almost double the size of the original. In all versions, a previously moralized task was fragmented and delegated to technical devices that prescribe action to numerous actors and objects (Riles Reference Riles2005). These instruments required judges, probation officers, and other practitioners to reorganize sentencing as a set of routinized and compartmentalized knowledge practices. Significantly, if these routines derive from histories of racist and classist dynamics, they could continuously enact these dynamics even without the involvement of actors’ beliefs.

Furthermore, the Federal Guidelines Manual ensured that all sentences emerged in prespecified data form, ready-made for collection and analysis. In this way, the internal composition of any future decision was guaranteed to conform to the commission’s specifications. With time, this immediate fit between decisions and data form became a property of the decisions themselves, and the labor invested in manufacturing this fit could fade into the background (Thévenot Reference Thévenot1984). When this occurred, commissioners, judges, and social scientists could simply “observe and measure” sentencing practices rather than actively commensurate them (Espeland and Stevens Reference Espeland and Stevens1998, 317). Finally, such instruments rearranged the external environment of sentencing by materially up-linking judges and decisions to an expansive data environment (Brayne Reference Brayne2017). A single decision thereby linked to every other decision in the country—past, present, and future—through a digital web of connective media. This was indeed a crucial phase in effectively performing the commission’s field panorama (Healy Reference Healy2015). The emergent data environment could now filter heterogeneous decisions through instruments already inscribed with a digital ontology that ordered them as data even before they were issued.

Boundaries Spanning

Lastly, boundaries spanning involves the enrollment of new groups of actors who could now plug into, and possibly affect, the continuous process of datafication. An important development in this regard was the public release of sentencing data by the USSC (1990, 28), which allowed numerous actors, including research agencies and academics, to utilize these data for various purposes (compare Lageson and Maruna Reference Lageson and Shadd2018).Footnote ⁷ As Jeffery Ulmer (Reference Ulmer2012) notes, when it comes to sentencing research, “[b]y far the biggest development in the late 1980s and 1990s was the availability of data collected by sentencing commissions in guideline states and the federal system.” This signaled a partial decentralization of sentencing practice and its redistribution among numerous actors “equipped with technical tools, who forge new types of expertise and plug into pre-existing ports in the body of the leviathan” (Eyal Reference Eyal2013b, 21).

By linking out to new actors and institutions, boundaries spanning expands the “data environment” or the “ecosystem of connective media” enacted by datafication (van Dijck Reference van Dijck2014). What allows for this expansion is the fact that datafication has already decontextualized and reformatted local practices into standardized forms. Importantly, while boundaries spanning may certainly augment the performative effects of datafication through enhancing digital or “dragnet surveillance” (Brayne Reference Brayne2018), it simultaneously makes its ideological underpinnings susceptible to contestation and change (Baack Reference Baack2015; Hannah-Moffat Reference Hannah-Moffat2019). At this stage, however, boundaries spanning had defined limits and directionality, largely due to the commission’s monopoly over collecting data and specifying its form (Ulmer Reference Ulmer2012). Even as data releases arguably made sentencing expertise less monopolized and more “generous” and “co-productive” (Eyal Reference Eyal2013a), all future analyses began with data whose structure and specifications were conditioned by an already fixed imaginary. As more actors utilized these data, they assumed the risk of further enacting the colorblind-carceral imaginary from which they derive.

Even agencies that utilized original data and were critical of the Guidelines could reproduce the commission’s panorama, as exemplified in a 1992 survey conducted by the Federal Judicial Center (1994). While the survey’s findings revealed judicial dissatisfaction with the Guidelines, its format enacted a vision of sentencing, judges, and the federal sentencing field that was not far removed from that of the commission. By envisioning judges as individual respondents with decontextualized opinions about the law, the survey embraced a perception of judges as separable data objects instead of law-making collectives; it cast judges’ “opinions” as discrete and measurable occurrences as opposed to what brings law into being; and it rendered the federal sentencing field as an aggregation of datafiable individuals as opposed to local organizations comprised of experts operating relationally. Though, in and of itself, the effects of a single survey are negligible, it joined a growing network of knowledge practices that imagine and enact judicial practice through a new panorama (Law Reference Law2009). While partially decentralizing datafication, boundaries spanning also diffused the new panorama—and, therefore, a certain vision of the judge, of sentencing practice, and of its temporal horizons—through space and time.

The Judge’s Eye in a New Panorama

That datafication enacted a new field panorama does not mean that it immediately homogenized judicial practice. In 1992, Circuit Judge Roger Miner (Reference Miner1992, 511; emphasis added) expressed the profound disjuncture between the new panorama and judges’ established ways of seeing:

Judge Miner simply could not “see” with this new set of optics. As another judge attested, this visual-cognitive failure also reflected a normative impasse: “It is simply not possible for judges to rigidly apply a general, two-dimensional set of sentencing standards to each individual case in a just and beneficial manner” (Weinstein Reference Weinstein1988, 11; see also Trott Reference Trott1995, 199; Wald Reference Wald, Sarat and Thomas1995, 84–86). As Jasanoff (Reference Jasanoff, Jasanoff and Kim2015, 32) notes, “imaginaries not only help reconfigure actors’ sense of possible spaces of action, but also their sense of rightness of action.”

This impasse inspired variable judicial responses, which we can broadly divide into three categories: collective-strategic resistance, individual tactics, and local differentiation by court communities. Initially, many judges invoked their established identity as a law-making collective or “the judicial branch.” By the end of 1989, two hundred judges had declared the Sentencing Guidelines unconstitutional, as compared to the 120 judges who upheld their constitutionality (Stith and Cabarnes Reference Stith and Cabarnes1998, 196).Footnote ⁸ Many also openly criticized the Guidelines in their decisions and other public forums (Cooper Reference Cooper1995). With time, however, judges relinquished the open strategic struggle to develop a new “common law of sentencing” (Berman Reference Berman2000) and employed covert tactics that straddled the old outlook and the new panorama (compare Ewick and Silbey Reference Ewick and Susan2003). Like other frontline actors with diminished capacity to shape formal law, judges found ingenious ways to circumvent the rules and manipulate the numbers (Brayne and Christin, forthcoming). What unites these tactics is an emphasis on the individual judge who experiences “an ethical dilemma” that inspires “creative sentencing” (Mihm Reference Mihm1997; Hillier Reference Hillier1991). Like other occasions wherein individuals avoid data surveillance, judges exercised agency by circumventing the Guidelines while creating the appearance of formal compliance (Nagel and Schulhofer Reference Nagel and Schulhofer1992). This, however, was not without its costs. As one judge expressed in an anonymous survey, “the guidelines … have made charlatans and dissemblers of us all. We spend our time plotting and scheming, bending and twisting, distorting and ignoring the law in an effort to achieve a just result” (Weinstein Reference Weinstein1992, 365).

A third response type involves variable absorption of the Guidelines by “court communities” (Dixon Reference Dixon1995; Ulmer Reference Ulmer2005; Wu and Spohn 2010; Lynch and Omori Reference Lynch and Marisa2014). While individual tactics remain isolated, court communities develop local “legal cultures,” which provide normative-cognitive cover (Church Reference Church1985). As studies have shown, local cultures partially account for the variable rates of “judge-initiated downward departures” from the Guidelines across districts (Kautt Reference Kautt2002; Lynch and Omori Reference Lynch and Marisa2014; Ulmer and Johnson Reference Ulmer and Johnson2017). To a significant degree, the gradual institutionalization of local norms accounts for the steady increase in national downward departure rates, as illustrated in Figure 7.

FIGURE 7.

Trends in departure status, fiscal year 1991 – fiscal year 2001.

Credit: USSC 2003, 32.

This brief overview of judicial responses demonstrates that datafication, even when pursued with vigor, cannot simply homogenize practice. Datafication is a continuous, dialectic achievement that is always in the making. To the extent that datafication assembles and enacts a coherent field, it operates alongside various local forces as a homogenizing lens that inserts itself into practice and relentlessly reformats recurring attempts to exit its scope. Thus, while creative sentencing and court communities reflect agency and creativity, such action is relentlessly rendered legible—indeed, captured—through the very vision to which it responds. While nothing can prevent judges from issuing sentences based on subjective considerations that have little to do with the Guidelines, these decisions will be continuously datafied, measured, and rendered governable. In this way, individual agency is partly prefigured through the very panorama to which it is ostensibly the antithesis (Tata Reference Tata2007). Similarly, “court communities” that variably adapt the Guidelines are made legible through the kind of large n analyses that the Guidelines themselves enabled. These communities can be seen as “local” by virtue of the capacity to statistically locate them against a “global” environment, itself enabled by datafication.

Continuous Datafication

From the perspective of the new panorama and the growing war on crime, however, the growing rates of “downward departures” became a burning political issue (M. Miller Reference Miller, Amin and Thrift2004). Congress responded by enacting the Feeney Amendment in April 2003, as part of the PROTECT Act, which instructed the commission to review the grounds of downward departures and take appropriate means.Footnote ⁹ The commission responded with alacrity. Following another extensive study of frequently cited departure reasons, it issued within months emergency Sentencing Guidelines amendments that reduced departure rates the following year (USSC 2003). Yet these formal amendments must be considered alongside humble and mundane adjustments in the daily flow of practice. The more profound and long-lasting effect of the Feeney Amendment was to further extend technologies of datafication and the data environment it assembles.

An important development in this respect involved the extension and elaboration of the commission’s data collection network. One amendment to the Guidelines, for example, pertained to an already-existing requirement from district judges to state the grounds for each downward departure “with specificity” in a written judgment and crucially in a data-collection instrument titled the statement of reasons (SOR) form (Figure 8).Footnote ¹⁰ Previously, such documentation was not mandatory and, importantly, was provided by probation officers and not the sentencing judge (USSC 2002, 39). The amendment shifted this responsibility to judges and thus embedded a digital trace in each and every decision. The broader significance of this mundane adjustment comes to view when considered alongside still other new requirements.

FIGURE 8.

SOR form (page 2/4).

Another provision required the chief judge of each district “to ensure that, within 30 days … the sentencing court submits to the Commission a written report of the sentence” and of various sentencing documents, including “the statement of reasons for the sentence imposed.”Footnote ¹¹ The Feeney Amendment further ensured that the commission “shall make available to the Attorney General, upon request, such data files as the Commission may assemble … including the identity of the sentencing judge.”Footnote ¹² Still other provisions instructed the commission to provide the amassed data to either Congress or the Department of Justice, thus establishing digital data collection conduits that linked every decision, as it was being written, to numerous actors in an expanding network.Footnote ¹³ These prosaic adjustments embedded a form of “dragnet surveillance” in the flow of practice (Brayne Reference Brayne2018), providing fast information, already in digital form, on every sentencing decision in an expanding network of knowledge practices. This effectively linked up “local” judges and practices to a “data environment” through which a host of “extra-field” influences could travel.

It is helpful to consider these changes in light of the four mechanisms of datafication. First, expanded dragnet surveillance increases the network’s density and its real-time capacity to translate heterogeneous decisions into a prespecified data form, thus relentlessly assembling local practices into a certain panorama. Second, the SOR form symbolically conveys the imaginary from which it derives by focusing attention on certain aspects of the task of sentencing and by being a persistent reminder to judges of their position within a data environment designed to measure and regulate their decisions. The SOR form thus connects each judicial chamber to other actors (the commission, Congress, social scientists, the public) within an expansive network. Furthermore, by making judges’ identifiers available in some instances, the judge-as-datum has become all the more visible and scalable. Figure 9, reproduced from the USSC’s (2012) Report on the Continuing Impact of United States v. Booker on Federal Sentencing, underscores this point. Each graph visualizes varying rates, by judge, of “non-government sponsored below range sentences” within a single district and all but invites observers (judges included) to measure each circle in relation to the norm.

FIGURE 9.

Reproduced from USSC’s 2012 Report on the Continuing Impact of United States v. Booker on Federal Sentencing: Part D (4 districts chosen at random).

Third, the SOR form also materially predisposes judicial practice—or “nudges” it, as Cass Sunstein and Richard Thaler (Reference Sunstein and Thaler2008) would have it—to conform with prespecified goals. Consonant to section 5K2.0(e), the Criminal Law Committee of the Judicial Conference designed the form to reflect the information that the commission needed to monitor sentencing practices (Gertner Reference Gertner2012). Yet the form’s design could enact the reality it purported to neutrally survey (Law Reference Law2009). First, the form requires the judge to report findings on the pre-sentence report and specify reasons for their modification as well as to elaborate findings on applicable mandatory minimum provisions. It then tracks the Guidelines’ categories: after instructing judges to report the applicable range calculation, it asks whether their decision falls within or departs from that range. In the latter case, the departure categories that the judge has to check are clearly enumerated. Thus, through embedded defaults and formulaic procedures, the form naturalizes a certain sentencing syntax and disposes practice regardless of judges’ orientations. In other words, the form can generate effects independent of the human agents who fill it out (Werth Reference Werth2019). As Steven Chanenson (Reference Chanenson2006; emphasis added) notes, “[i]ronically described as a ‘Statement of Reasons,’ … the form almost seems designed to encourage the kind of mechanical—and arguably unreasoned—approach to sentencing Booker tried to extinguish.” A judge who tries to exit this vision by using the open-ended space to “explain the facts justifying the sentencing outside the range” must realize that her idiosyncratic reading would collapse back into available categories and be consequently plotted in a graph. The cumulative effects of these operations allow us to appreciate the impact of the mechanisms considered thus far. Through the SOR form and related operations, sentencing decisions could be reformatted into data and, in this form, uploaded to an expansive network in ways that can be grafted onto how others are seeing, thus linking them to the network and eliciting their cooperation.

Lastly, this elaborate and digitally embedded reformatting process leads us to boundaries spanning. The Transactional Records Access Clearinghouse (TRAC) offers a helpful illustration. TRAC operates as part of Syracuse University with the mission to obtain case-by-case records of all federal criminal cases and make their analyses accessible to the public.Footnote ¹⁴ The purpose of TRAC, according to its directors, is to study “how well sentencing practices were working and to … develop detailed data as well as web-based tools that would allow the public as well as the courts to examine these sentencing practices and monitor it over time.”Footnote ¹⁵ Some of the tools available on TRAC’s website allow users to investigate the sentencing averages of federal judges, by name, and measure them against other judges, their district, and the entire federal judiciary (Figure 10).

FIGURE 10.

Cropped image from an average prison sentence report for a presiding judge for the Middle District of Florida.

Credit: TRAC website, https://trac.syr.edu/judges/aboutData.html.

Judges, in turn, are well aware of the capacity of dragnet surveillance and have become mindful of the analyses that TRAC and TRAC users undertake (Gertner Reference Gertner2012). As the New York Times reported, “[t]he trove of data subjects individual district court judges to a level of scrutiny unprecedented in the history of the judiciary. … In the Eastern District of New York, for example, the 28 judges in the study delivered a median sentence of 24 months for drug cases in the past five years. But there were disparities: Judges Jack B. Weinstein and Kiyo A. Matsumoto gave median drug sentences of 12 months, while the median drug sentence for Judge Arthur D. Spatt was 64 months” (Secret Reference Secret2012). In the reality enacted pursuant to the Feeney Amendment, the pathways between data analyses and judicial practice have become well established. A judge reaches a sentencing decision, already encoded in a data form prescribed by the commission through the SOR form; the decision is then uploaded onto established data trails and travels to the commission; the commission compiles the data and makes them available, if modified, to the Attorney General, Congress, and researchers; the commission and increasingly other actors analyze the data and circulate their findings within and without the very field they describe; when needed, the commission issues modifications in the SOR form and/or amends the data-collection network on which its work depends; gradually, numerous other actors rely upon these data to monitor and analyze sentencing practice.