Carceral Data: The Limits of Transparency-as-Accountability in Carceral Data: The Limits of Transparency-as-Accountability in Prison Risk Data Prison Risk Data

Prison data collection is a labyrinthine infrastructure. This article engages with debates around the political potentials and limitations of transparency as a form of “accountability,” specifically as it relates to carceral management and data gathering. We examine the use of OASys, a widely used risk assessment tool in the British prison system, in order to demonstrate how transparency operates as a means of legitimating prison data collection and ensuing penal management. Prisoner options to resist their file, or “data double,” in this context are considered and the decisive role of OASys as an immediately operationalized technical structure is outlined. We demonstrate that the political and managerial logics that underpin OASys heavily shape and structurally limit the terrain on which any individual might challenge their carceral administration. Ultimately, we argue that, in this context, transparency itself forecloses broader notions of radical and systemic change by inviting individuals into the creation of their own data double, and thus the legitimation of their own penal management and the forms of procedural justice offered by the prison estate. Finally, implications from this case for broader debates about transparency-as-accountability are explored.

accountability, criminal justice, OASys, prisons, risk assessment, secrecy, surveillance, transparency This article investigates the operative role of risk assessments and data collection in British prisons. Through this case study, we examine important limits to calls for data "transparency" and "accountability" when it comes to substantively challenging the power dynamics of carceral datagathering practices. The piece considers risk assessments, specifically the Offender Assessment System (OASys), as a key tool in the data infrastructure of British prisons. We interrogate how "data subjects" (Goriunova 2019; UK Information Commissioner's Office 2022) or "data doubles" (Bouk 2017;Haggerty and Ericson 2005;Raley 2013) are produced as mechanisms for directly monitoring and managing "offender behavior," and shaping an individual's trajectory through the prison estate and into the community. Such assessment structures form an essential vehicle through which people imprisoned for particular offense types are controlled within and after prison, and become central in characterizing penal supervision at scale (Howard and Dixon 2012). This landscape presents a particular challenge to arguments around the importance of data transparency and accountability (Diakopoulos 2016;Laat 2018;Pasquale 2015), and the ways in which these appeals might inform the campaigning activities of activists, prisoners and their loved ones, as well as interested scholars. Is the prison file a productive terrain on which to argue for transparency or data accuracy? How might calls to reveal more data inform subject only insofar as it directly informs discussions around risk, behavior management, and control (Deleuze 1992). We therefore argue that the creation of data subjects through technical structures such as OASys, so directly tied to issues of power and management in prison, can inform broader questions about data transparency and accountability. We highlight the limits of transparency as a framework for addressing injustice as it tends toward a procedurally restricted horizon of accountability, and legitimizes structures inherently tied to practices of apprehension and control.
We understand transparency as a necessarily contested concept (Ananny and Crawford 2018). It is one that is both mobilized in academic and popular discussions for progressive reform aimed at increasing the responsiveness, accountability and democratic functioning of institutions (Diakopoulos 2016;Laat 2018;Pasquale 2015), as well as being adopted and deployed by government organizations (Ruppert 2015). As concepts are adopted and embedded into institutional policy and practice, their meaning can be polyvalent (Dobson 2022) -differing across civil society critique, institutional actors' understanding, workers' practice and "service users'' experience. For the purposes of this article, we consider transparency as it is invoked within the literature that scaffolds prison risk assessment: in policies, guidance and, ultimately, in assessments' logic. It is this transparency, itself adopted by prison procedure, that we hold up against academic and popular calls for greater transparency. We ask what this might mean. A full accounting of the gaps between those calls and transparency's "real" operation in prison is beyond the scope of this article. Yet, in positing how transparency operates to legitimate forms of penal management, we point to one such gap.
Equally, though we refer to and have conducted empirical work on how transparency operates in quotidian practice in prison, this paper focuses specifically on the operationalization of transparency-as-accountability in assessments' logic, structure, and professional literature. As such, it is the architecture of prison risk assessment that forms the case study of this article, and it is this to which we refer throughout. Though prison presents an acutely constrained environment for assessments to be resisted, contested and sidelined, our analysis does not preclude the potential for alternate approaches to assessment. Its focus, however, remains on assessments' embedded logic, and the guidance that scaffolds this logic.
Our case study is OASys, a computer-based tool used to assess imprisoned individuals and inform their sentence plans. OASys is utilized to assess "risk" on a number of scales and, perhaps most crucially, predict recidivism within and outside of prison. For the prison estate, as a network of institutions across England and Wales, such standardized forms of assessment both facilitate the management of offenders in institutionalized terms (making one institutional or professional judgment immediately legible and operational to the next), but also form a constitutive part of the With these questions, we take OASys as a case study through which to interrogate the operation of data transparency and accountability.
Ultimately, we argue that the data subject produced through assessments like OASys is so imbued with the political and managerial function of data in the prison context that avenues of transparency and accountability are in fact part of the way such data subjects operate. That, in prison, we see transparency form an important role in how such data systems legitimate their own processes, function and their exertion of control and management over the people represented in data doubles. Transparency and accountability then, need radically rethinking in contexts wherein the logics of data collection, operationalization, and redress are imbued heavily with the political logics of the institutions that gather, disclose, and offer transparency-as-accountability.

Transparency and Opacity with Data
There is a common sense notion of transparency that understands it as an inherent good and necessarily generative of "accountability." Much popular discourse, as well as specialist and academic calls for data parity (Ada Lovelace Institute 2020; Blacklaws 2018), rely on the assumption that by building in frameworks or developing tools for transparency, we are better able to hold systems to account (Diakopoulos 2016). Underlying, for example, the idea that we live in a "black box society" (Pasquale 2015) is a notion that, by opening up the black box, and seeing inside the system, we might be able to access some insight and truth of that system's operation.
So, it follows that by accessing such insight, we may be better positioned to hold these systems to account (Reggi and Dawes 2016), access forms of redress (Armstrong 2014), and bring their workings into public view for discussion and challenge (Blacklaws 2018). As Ananny and Crawford (2018, 974) have articulated, underlying this notion of "transparency-asaccountability" is the belief that "seeing a phenomenon creates opportunities and obligations to make it accountable and thus to change it." There is an emergent critical literature that seeks to challenge these preconceptions around transparency, particularly concerning the proliferation of data (Aradau 2017;Birchall 2016;Ruppert 2015;Valdivia et al. 2022). In these accounts, the forms of accountability transparency offers cannot be delivered by transparency itself. Extending this literature, this article outlines how forms of institutionally offered transparency not only fail to address structural injustice, but actively contribute to limiting frameworks of redress, foreclosing political opportunities, and legitimating forms of data collection and ensuing control. To put this another way, we argue that transparency-as-accountability not only presents a limiting and, at times, obfuscating horizon but that transparency and its linking with accountability itself acts to legitimate and obstruct challenges to systemic injustice.
Opacity has been understood to refer not "to hiddenness, but to the incomprehensible or inaccessible ways in which systems work" (Valdivia et al 2022, 3) as opposed to the mere limitations of access associated with secrecy. Certainly prison's risk assessment and management architecture is overwhelming in its labyrinthine complexity, replete with acronyms. Indeed, this article can only describe a fraction of its architecture. As discussed later, however, transparency is invoked in prison risk assessment guidance, training and policy as a direct counter to these recognized forms of secrecy and opacity. Transparency is put forward as a limit and challenge to prison secrecy and opacity, inviting prisoners into the construction of data about themselves in an open and collaborative process, wherein decisions about them and interpretations of their behavior can be said to have been coconstructed with prisoners' involvement. Recognizing the importance of this move, we turn instead to obfuscation in the assessment process. We trace the manner in which this proudly transparent, collaborative process partially precludes, forecloses and otherwise muddies other routes of challenge that prisoners and their loved ones might pursue when trying to confront prison decisions about their sentence, treatment, and their character.
Transparency's relationship to opacity, here, is central. For example, Valdivia et al. (2022) have shown how the binary between transparency and opacity often restricts forms of both analysis and redress when examining modes of "datafication." Meanwhile, Fuller (2016) argued that transparency is a way of producing certain occlusions of transparency at the same time.
For him, transparency is a kind of "parallax effect" in that what is transparent to a human differs from what is transparent to a machine (Fuller 2016).
Scholars have also outlined the political implications of this duality between transparency and secrecy. For Birchall (2011), "transparency" and "secrecy" cannot be separated. However, she argues that, politically, this is not an inherent problem for radical change, as the "aporia of transparencyas-secrecy and secrecy-as-transparency" becomes an opportunity to operate in this contradiction, arguing for "transparency qua secrecy, secrecy in transparency" (Birchall 2011, 13). The kinds of opportunities and operations of power folded into the construction of secrets may offer avenues for intervention and provocation that may be politically fruitful (Brabazon 2022).
How in prison, then, does the relationship of transparency (as itself a means of opacity) operate? Literature on prison management of information highlights the clearly outlined avenues for redress, disclosure, and means of dialogue that prison offers those kept within its walls, as well as their loved ones (Armstrong 2014). Prisons thus remain ready for requests for disclosure, as this operates within the institution's restrictive framework of "procedural justice" (Wells 2008). These might play out managerial forms of accountability that are demonstrable within such organizations' "audit culture" (Shore and Wright 2015), but fall short of opening up to transformational change in any broader, systemic sense.
To ground transparency as it operates in relation to OASys, we must consider the importance of the data subject as something produced through, and monitored by, such assessments. A central element of much literature on data concerns the construction of a "data subject" through the collection of information about individuals at scale. For Amoore and de Goede (2008), data collection is central to risk-based calculative logics that get deployed to manage both populations and individuals. Bouk outlines the history of this relationship and argues that, in the 1970s, personal data "shifted toward the use of aggregates to fit individuals into the mass" but that we are now in a conjuncture in which "individuals generate many data doubles that are commodified, capitalized, collected, celebrated, and often out of the control of those they represent" (Bouk 2017, 85). The "data double" outlined by The relationship of this data subject to a concretized or "real" individual is at best ambivalent (Goriunova 2019, 126 (Werth 2019b) and rational ways from "truths" foregrounded through the construction of data subjects, does getting to see and dispute this truth matter? Surely revealing and challenging the accuracy of one's file offers one route to substantive change.

Data Governance and OASys
Prison data collection is a labyrinthine infrastructure. Upon entry to the prison estate, one already has significant data baggage that shapes the forms of assessment to which one later might be subjected. The type and details of one's offense, a judge's sentencing remarks, victim statements, pre-sentence reports and one's history of contact with the criminal justice system all shape entry and induction into prison before the prison's data collection commences in earnest.
Piloted in 2001 and later rolled out across the prison estate of England and Wales, OASys has a standardized assessment format that seeks to capture the risks and needs of people under custodial supervision. 3 The OASys, held across computer systems and integrated and updated electronically, allows prison systems to "speak to each other" through their interoperable form. 4 Importantly, OASys assessments are not completed for every prisoner, but their use is standard for those who have committed violent or sexual offenses, have a sentence of over a certain number of years, or those who have been convicted under specific laws or offense types.
An OASys assessment has a number of sections which are completed for all those who are assessed, as well as specialized sections that might be completed depending on someone's specific offense. OASys is one assessment amongst many. Someone may be assessed with OASys as well as a specific tool for violence, sexual offending, substance misuse, or a psychological assessment in full prose, report form. However, of the suite of 3 The system's abbreviation, OASys, pronounced "oasis," may strike readers as deeply ironic as a surveillance tool within prison with a name associated with spaces of peace, respite and hope. It is perhaps worth noting that the Ministry of Justice operates using innumerable other acronyms, such as ASPIRE (Assess, Sentence Plan, Implement, Review, Evaluate), ARMS (Active Risk Management System), and CARAT (Counselling, Assessment, Referral, Advice, Throughcare) and many more. All these acronyms exhibit an at times unsettling culture of jargon that neutralizes and distances staff and policy from the day to day experiences of imprisonment.

4
Prior to the OASys, other forms of assessment and reporting were used to formulate understandings of risk, case management and prisoners' needs and difficulties. Presentence reports, which professionals were trained to complete, were far shorter than an OASys assessment -often just a few pages -that summarized the prisoners' offense, situation, difficulties and needs and the implications of this on managing a prisoner's risk. Largely superseded by the OASys in the early 2000s, justice workers are now invited to record a much larger set of data to comply with the OASys' sections covering multiple areas of risk. detailed assessments conducted in the prison estate, OASys is the most common. 5 The language OASys uses and generates is both quantitative and qualitative. Those completing an assessment are encouraged to identify existing needs, risks, and issues present for the person whom they are assessing, and are asked both for numbered scores about that person's risk in each area, as well as encouraged to give a longer, qualitative explanation of why and how someone's risk, needs or other issues present in the manner that they do. Risk scores are generated both for an individual's perceived risk to different groups, as well as to themselves, and for a set of criminogenic needs present in an individual, that range from thinking skills to relationships to substance misuse and financial management. 6 The tool is specifically designed to assess and reduce recidivism, and to indicate interventions most likely to contribute to reducing someone's perceived risk.

5
It is important to note, however, that OASys is not conducted with all prisoners, but usually reserved for young offenders, those with serious offenses, longer sentences, those sentenced under certain laws and those who have been recalled to prison. The tool's promise to facilitate risk management, here, becomes crucial in terms of its ability to divine the criminogenic needs, risks and resources required by a prisoner in one of these categories -those with longer sentences, serious offenses. By OASys' own description, the assessment is designed to: "assess how likely an offender is to be reconvicted; identify and classify offending related needs; assess risk of harm (to self and others); assist with management of risk of harm; link assessments, supervision and sentence plans; indicate any need for further specialist assessments; measure how an offender changes during the period of supervision/sentence." 6 Perceived risk to self, here, refers to the estimated likelihood that someone will engage in self-injurious or suicidal behavior. Assessments regularly list the number of "ACCTs" (Assessment, Care and Custody and Teamwork) someone has been given. ACCTs are plans of how to watch and support people at risk of self-harm or suicide in prison. Having many ACCTs thus may indicate someone's heightened 'riskiness' in terms of their propensity to act on feelings of distress. Particular scores within the assessment correspond to thresholds of perceived risk. Scores of "High," "Low," or "Medium" risk are generated towards the end of the assessment. An individual's score along these risk scales goes on to inform plans and interventions involved in decisions about one's sentence, rehabilitative courses to be undertaken, transfers in prison, and day-to-day behavior management.
The implications of one's OASys are thus broad. The assessment is often conducted at points of change within one's time in prison: at induction, recall, before parole, or after a significant change in one's circumstances in prison. Its scores and analysis are designed to be immediately operationalto inform the kinds of courses, programs, and placement of an individual within custody, to help professionals decide whether to release, move or extend the sentence of an individual, and to decide license conditions within the community upon release. Indeed, the tool is designed to lead professionals through consideration of how to construct a risk management and sentence plan that targets the areas of criminogenic need from which an individual's risk and offending behavior is seen to emerge. For practitioners working with imprisoned people, such assessments are thus imbued with an ability to obtain, decipher and record crucially important information about an individual that has immediate practical application in terms of institutional decision-making. The data extracted and stored through OASys, then, is immediately mobilized.
Importantly, scores are generated through a process of comparing like-for-like prisoners using OASys' own historic data and wider recidivism databases. For example, an individual's score on the "Risk of Serious Harm" section is generated by comparing data gathered on this individual to a database of information about other, comparable individuals and their predicted and historic records of causing what is categorized as serious harm. Data subjects generated by OASys as a whole system across the prison estate, then, are put in statistical dialogue with one another as they are asked to generate scores that shape individuals' trajectory in prison. For prison professionals, OASys is rarely a singular tool to algorithmically divine what should happen to a person in prison, or at parole. But it is through OASys that specific needs are monitored, gleaned and brought to the fore.
Justice professionals' work often involves taking this information to guide their discussions on interventions, transfers, treatment, and next steps (Maguire and Raynor 2017;Wendy Fitzgibbon 2008). Furthermore, OASys' standardized, interoperable, and technical nature also makes the assessment easily replicable. Professionals can include prior information about someone's index offense, behavior in prison, and previous lifestyles and skills, for example, from previous OASys forms. 7 Information input into 7 Each new assessment produced across one's time in prison, though often automatically drawing on data from those before it, folds in these new pieces of information to inform the next set of decisions that present themselves to professionals tasked with offender management, treatment plans, sentence progression, release decisions, license conditions or risk management plans. In sum, the edifice of data collection and OASys is, therefore, often multiplied and reproduced in future OASys assessments, generating a recursive, layered charting of an individual's time in prison. For many, the foregrounding or misunderstanding of particular, perhaps otherwise negligible incidents, can thus take on a significant role in their OASys as it is faithfully reproduced by the data system year after year.
Here we see the construction of a data subject through OASys. It is easy to imagine, considering the importance of this data subject in shaping one's trajectory in prison, how disputes over the accuracy and, as is more common, the meaning of information collected through assessment might arise. 8 What remains acutely clear is that any data subject generated by an OASys is produced with a particular political and managerial purpose.
Criminogenic need, beliefs, activities, and behaviors are sought out and directly related to offending behavior. It is these kinds of information, interrelated in these kinds of ways, which are elicited and sewn together application within prisons through assessments is a labyrinthine set of ever-changing and heterogenous tools and techniques.

8
OASys can contain things that are factually wrong, such as someone being alleged to have done or said something that simply did not happen. Often, however, legalistic disputes center more on the particular meaning or stated significance of something someone did indeed do or say. For example, the weight or intended meaning of a comment using diminutive language -calling an officer "pet" or "love" -could be disputed. In some cases, this would not be considered significant. If an individual was understood to hold demeaning views about women as part of the beliefs that supported their offending behavior, however, it may be considered highly significant. If their relationship to authority was understood to be particularly fraught and closely linked to their offending, too, such an incident might take on particular significance. If recorded by staff and interpreted as meaningful and relevant, then, an incident like this could be repeated in many OASys assessments and become an area of important dispute as to the weight and meaning of such a comment. Where a comment like this was taken to be significant, an imprisoned person might be expected to articulate how and why they were working on changing their language, and attitudes to women, or to authority figures. through an OASys. Here political and institutional notions of criminality are given "categorical foundations" in the very structures of data and assessment (McQuillan 2015, 566). Through this process, a specific account of the individual is produced. As such, OASys produces a "data subject" or "data double" that becomes the operative object to manage and control through prison interventions. These record-keeping systems and risk assessment tools thus take on a decisively active role through which people are measured against one another, then shaping the pathway available for anyone. Here, then, we see how particular aspects of the individual are brought to the fore through technical processes of data construction and stabilization, and that these aspects go on to inform methods of management at greater and greater scale. The data double that is constructed may have varying degrees of accuracy or fidelity to any individual. Its closeness to someone's own understanding of their history, motivations, struggles and needs is ambivalent, but its tethering to particular imperatives of the prison is evident. Risk assessments directly shape how one is managed in prison.
Assessments are understood to be backed up by an edifice of research on their ability to divine recidivism, identify core issues for an offender and contribute to clear plans and comparative scores by which professionals can manage and monitor individuals (Werth 2019a). They produce a data double which clearly matches the goals of the penal estate. It is this form that must be acted upon, because it is through these assessments that the kinds of needs and challenges that relate directly to the institution's goals, and its staff's training and objectives, are matched.
This operates both at an individual scale, in shaping discussions and decisions made about a person, but also en masse. At an institutional scale, this mass effort of data collection, record keeping, and application comes to constitute the prison's logic more widely. As such, the transparency and opacity of prison data collection is central to the experience of tens of thousands of people. Through OASys as a whole, an edifice of comparative data about criminals, crime and imprisonment across the prison estate are interrelated and operationalized. Accuracy and fidelity to "real subjects" becomes even blurrier at this critical mass, wherein the managerial promise of OASys as something tied to and supportive of the prison estate's core objectives imbues assessments with its logic from the start.
The data subject thus scaffolds the parameters and options for professional decision-making around an individual: their sentence, their progression through the prison system, what services may be made available to them, and what programs they might be required or expected to complete. Yet more than this, at scale, data subjects inform the edifice of risk estimation and ensuing decision-making across the penal estate. Their data constitutes part of the meters of risk that determine how other prisoners' data is interpreted, and feeds into a massified set of data subjects that present a picture of risk across the penal estate. In turn, these data are fed into research that swells the evidence base undergirding the importance and usefulness of risk assessment. Each of these layers is matched neatly to the imperatives of imprisonment itself. These data subjects thus continually set and reset, in a feedback loop, the parameters of understanding and decision-making about prisoners at individual and institutional scales.
There are a number of ways in which transparency forms a constituent part of this process and is a tool by which the prison makes claims towards accountability. Here, that underlying political and managerial logic cannot be disarticulated. In the intensified forms of management and control embedded throughout the prison system, the data subject takes on a particularly decisive and alienating form: it searches for particular kinds of answers, has decisive power, forms a yardstick by which one's development -and others' risk -may then be judged, and comes to shape one's reality, in terms of what one can and cannot do, in assertive ways from the moment the data double is created.

Legislating Transparency
Access to one's data within prison is managed by a set of legal frameworks that emerged in response to personal data and disclosure issues. Within legal and legislative apparatus, the notion of a data subject generates a certain economy of transparency as articulated through an 20 Submission to Secrecy and Society https://scholarworks.sjsu.edu/secrecyandsociety DOI: 10.55917/2377-6188.1074 institution. This mode of transparency-as-accountability is organized around balancing the demands of an information-collecting institution on the one hand, and the subject of collection on the other. The differentials of power and knowledge in these situations are rarely adequately mitigated through policies of transparency policies (Crewe 2011). In the prison, the direct imperatives of penal management, control and behavioral monitoring are taken as paramount to individual data rights. Despite stark asymmetries and institutional imperatives, policy language often collapses a "data subject" and an institution into the language of equal participants in a contract, implying an even playing field.
Making personal data transparent and available to the subject of data collection has a particular political and legislative history. It is a history in which the subject is first recorded, analyzed, researched, and surveilled by state and non-state organizations, and then increasingly receives the right to access this information. This, too, entrenches the notion of parity. As Vismann (2008, 148) argues, the right to information is often framed as a "subjective right to inspect one's own records" in which the "social demand for public transparency is transformed into an individual's claim to valuable assets." Here the institutional necessity for secrecy is measured against the individual demand to be informed, producing a situation in which "the interests of both sides appear to be equal, making it possible to weigh the pros and cons" (Vismann 2008, 148).
Since the beginning of the twentieth century, we have seen in the UK a number of key legislative frameworks around granting access to data collected on oneself. Broadly, these have been incremental reforms towards what gets characterized as a more open or transparent democracy. There have been several key interventions regarding the right to access information an organization has collected on you (Bartlett and Everett 2017). Currently, the right to one's data is governed by two frameworks: the Data Protection Act 2018 (DPA) and UK General Data Protection Regulation (UKGDPR). Under this legislation, an individual has the right to request access to the personal data an institution has on them, including information collected by government departments. Together these frameworks aim to safeguard personal data, balancing the needs of organizations to obtain and utilize personal data with the rights of individuals to privacy (UK Information Commissioner's Office 2018). Here, the language of rights and property help to scaffold principles of transparency and disclosure.

Indeed, within the framework of the Data Protection Act and UK
General Data Protection Regulation (GDPR), a limited and specific notion of the "data subject" emerges and is conferred particular rights over their information. Within the legislation, a "data subject" is someone identified from personal data; "the data could be their name, address, telephone number or something else -but if it's about a person, then they're the data subject. They're the 'subject' of the data" (UK Information Commissioner's Office 2022). This administrative category also manages and governs the relation of information between the individual and the institution collecting the data (what gets called the "data controller" in GDPR terminology). It is a mode of individuation that grants rights only to data pertaining to you as a subject. Here we can see how the accountability this framework offers dovetails tightly with the right to see one's data, to access transparency as accountability. This accountability is, too, highly individualized. The "right for the protection of 'personal data' that is in some respects reminiscent of a claim of ownership" (Vismann 2008, 149). The deployment of assessment structures and the operationalization of transparency-as-accountability within the prison enables modes of institutional decision-making that appear to be internally consistent, neutral, impartial, and, therefore, just and fair. Within the criminal justice system, this is articulated as a form of "procedural justice," a notion that wields a significant influence over prison policies and practices. As Wells (2008) has argued, "techno-fixes" within the criminal justice system that aim to identify risk, such as OASys, are presumed to be unbiased and reliable mechanisms and, therefore, are seen as procedurally just. At an institutional level, however, we can see here how procedural forms of justice set out to both placate those aggrieved by the prison system and produce a form of accountability organized around the belief that the process was just. This is highlighted in a recent Ministry of Justice and HM Prison & Probation Service policy framework in which they state that when people "believe the process of applying rules (how a decision is made rather than what decision is made, and how they are treated during the process) is fair, it influences their views and behaviour." It goes on to emphasize this point, "people are much more

Transparent Subjects and Built-in Transparency
Although prison systems tightly restrict access to data concerning offenders, OASys is not concealed from the person the assessment The notion of transparency is intrinsic to how OASys is understood, conducted, and its contents deemed credible. Indeed, an individual being assessed is brought into the process of making their data double. OASys assessments are conducted with prisoners and OASys guidance dictates that prisoners are an essential part of any assessment (HM Prison Service 2005) and that their inclusion and a principle of "openness" guide assessment interactions. The ability to conduct a constructive discussion with an individual through the assessment tool is a core principle of the assessment's guidance. Discussion with an individual about their life, attitudes and offending is encouraged (HM Prison Service 2005). This is highlighted by a portion of the assessment asking individuals to complete the self-assessment questionnaire to ascertain their own estimations of their needs and risk.
As outlined in guides and training for justice professionals, an OASys necessitates engagement and transparency: it is itself an offer of transparency about the prison's decision-making, and an opportunity for participation in constructing one's data double. Through channels scaffolded by OASys itself, one is invited to participate in this construction that will go on to shape the kinds of treatment, behavior management and monitoring that will flow from this data double: one's own penal risk management becomes an arena in which one is invited to participate. Therefore, a particularly delineated form of transparency is built into OASys as part of its very function. As an organizing principle of its governing policies, transparency looms large in OASys. Notions of accountability, as well as prisoner compliance, undergird this. It is understood that the depth of "engagement" a prisoner has with their own assessment and treatment is an important indicator of willingness to reduce risk, and reduced recidivism (Maguire and Raynor 2016). Moments like an OASys assessment, then, on top of the various risk reduction programs that might be recommended to an imprisoned person to engage with rehabilitation, is itself a prime opportunity to "engage" an imprisoned person, as well for the imprisoned person to demonstrate their own engagement.
At those decisive points of disclosure, too, accountability to the imprisoned person is demonstrated by this institutional transparency. Access to one's OASys report arises at key moments of procedural decision-making, organized around the legal and penal imperatives that govern procedures such as Parole Board hearings or challenges to one's case. The legalistic structures of "transparency," as this opens up predetermined avenues of engagement, openness and dispute within the penal system, are thus welldeveloped parts of OASys' everyday functioning and the policy principles that govern its use.
Here, transparency's interrelation with forms of obfuscation and procedural justice comes to the fore. In working along principles of openness and transparency, OASys offers given channels through which imprisoned people come to participate in forms of their own assessment and management. As Foucault addresses when discussing prison "reforms" in 1976, reforms often involve producing a situation in which incarcerated individuals are made to "participate in devising the very programs for their punishment" and accept and internalize the punitive procedure by acknowledging culpability (Foucault 2009, 16). Disclosures of OASys data, too, continue this form of transparency as a constitutive part of punitive control. Once one sees the contents of one's OASys: what it recommends in terms of needs and risks to be managed, one is further invited into the construction and its ensuing operationalization.
Transparency thus operates not as a form of mere opacity -making unintelligible or inaccessible the logics of prison decision-making, risk assessment and ensuing management of prisoners -but as a means of obfuscating other avenues of dispute. One has been recruited into the construction of one's data double through, as per best practice guidance, a transparent and collaborative interview and self-reporting process. Open engagement with this process lays the groundwork for a more 'accurate' and detailed assessment of one's needs, and one's compliance with systems of prison risk management. Contesting, refusing or rejecting the risk assessment process is thus heavily disincentivized. Post-assessment contestation, too, loses its edge: you were involved in constructing this data double, it incorporates your thoughts, reflections and self-assessment surveys in its reading. The ensuing characterization of one's risks, needs, sentence and treatment plans are made plain through completed assessments, openly shared with the individuals concerned. This serves as one's plan going forward: this is the way through, and the way out. This process, from the construction of one's data double to faithful engagement with its recommendations, lays the parameters of one's progression through a sentence and out of the custodial estate. In incorporating transparency into its operation, OASys thus obfuscates other avenues of dispute or redress and draws imprisoned people into the production of their own data double, and the logics of penal management generated from this double.

Transparency and Penal Management
Perhaps most importantly and most evidently in the prison setting, overtures of transparency invite imprisoned people into their own behavioral management. OASys' logic, drawing out particular elements of an individual to be assessed and to constitute a data double whose risk can be monitored and acted upon, is also "built in" with transparency. Its very form is tethered to the needs of penal management, and its incorporation of transparency cannot be disarticulated from this purpose. The practicable import of data collected and constructed through OASys, and similar assessments, is clear: they are designed to be used, and are immediately enmeshed in a set of decisions prison staff must make about those in custody. These tools harvest information to divine potential future behavior of individuals, and inform institutional strategies, sentence plans, resource allocation and the activities of justice workers around an individual's case. The prison is thus always mobilizing data constructed through tools like OASys as part of its operation and activities. Though assessments may only take place sporadically, their import in charting someone's course through the prison's behavior management, support programs, treatment offers, and other "risk reduction" work is perennially active.
OASys' logic constructs specific aspects of an individual as visible, urgent, intelligible, and computable. Its foregrounding of "criminogenic need" means that certain areas of a person's life -those scoring particularly high, and those understood to be most closely related to offending behavior -become visible to the system and urgent to manage. OASys matches these needs to the prison's ability to monitor, intervene and act on these issues. The suite of offender behavior management courses, therapeutic interventions and different prison environments designed for different offender needs are sought to match those outstanding risks and needs.
These include specific therapeutic prison wings such as Therapeutic Communities or Psychologically Informed Planned Environments, substance misuse services, in-cell workbooks on anything from attitudes to offending to processing emotions, offender behavior courses ranging from drink driving awareness to general "Thinking and Skills" programs to intensive courses on Healthy Sex and Relationships or uses of violence. Particular kinds of administrative, therapeutic, and behavioral interventions are thus matched and targeted at those areas that have been surfaced as the most linked to risk through the OASys. The areas that are considered "core risk reduction" work, matched to the needs surfaced through an OASys, must be engaged with to a satisfactory level, completed, and their impact evidenced, before one's commitment to and success in risk reduction may be favorably measured at the next juncture. An individual's willingness to productively and earnestly engage with these interventions becomes central to later assessments and estimations of risk and change. The potentially decisive surveillance of the next OASys is thus always integrated into everyday behavior management, and the objectives of risk reduction interventions.
Depending on the kinds of intervention recommended, demonstrating reduced risk might mean showing changed behavior in terms of responses to others and activities but, more often than not, also includes work to act on beliefs, thinking processes and attitudes that are understood to constitute one's risk and offending. Being able to demonstrate insight into one's beliefs, lifestyle, skills, activities and more as contributing to one's offending -and therefore in need of reform -is part of progressing through a risk reduction journey. Here, the germ of the data double generated in assessment produces a kind of "script" that someone must learn, adopt, and enact. Information structures, then, not only record the individual in specific ways, and read particular behaviors as meaningful or worthy of attention, but also form a decisive, constitutive part of the infrastructure that coerces individuals into acting in ways that align with the intended rehabilitation pathways set out through initial assessments. Actively complying and demonstrating that one accepts, internalizes, and works on the construction of self as produced through OASys can become the path of least resistance get through a period of imprisonment.
As an illustrative example, once particular attitudes, beliefs, behavior, and needs are brought to the fore of the prison's conception of someone in their data double, behavior from then on is often read back through this frame. An otherwise innocuous comment, altercation, indulgence, interest or lack thereof might raise alarms because it matches onto identified risk factors. This can, if recorded by prison staff, be integrated into the next OASys or comparable risk assessment. If social isolation, for example, has been identified as a risk factor through assessment, then deciding to stay in one's cell, refusing to engage in group activities, or an increase in quiet, reclusive behavior can raise alarms about an emphasized criminogenic need.
Such assessments, therefore, are productive in that they generate feedback loops and a recursive logic. For a person to improve the estimation of their risk and progress through prison, they must reduce their risk in ways that are measurable and intelligible to the technical and administrative structure of OASys. Personal change in terms of a reduction of risk, then, is both constituted through an OASys and needs to be legible to the next OASys. Here, engagement again becomes paramount. One's earnest involvement with the therapeutic and behavioral interventions offered, as well as things like skills training, jobs within the prison and day-to-day behavior may feed into future assessments. Opportunities to progress are thus scaffolded by the kinds of data constituted in one's data double, and improvement needs to be demonstrable and legible to these forms of data collection in order to be recorded and actionable.
An individual's data is fed into systems that predict and manage their behavior. Whether such doubles are transparently disclosed, riven with errors or reflect accurately the experience of a subject, their political functionality in managing and conditioning behavior as part of penal control is paramount. The OASys becomes subject producing, insofar as institutions respond to its constructions and begin to act on those individuals represented by data doubles. For imprisoned people, involvement in and disclosures of one's OASys are part of this process; one is invited to participate and shape a data double and, once one learns the risks, needs and areas for improvement surfaced in this double, it might condition how one behaves and speaks, what one engages with, and the kinds of activities, therapeutic work, and programs in which one participates to facilitate progress through prison.

Legitimating with Transparency-as-Accountability
So how does transparency-as-accountability operate here? Having traced OASys' practical implementation, we can see that transparency within assessment works to undergird the assessment's own legitimacy and operationalization. Openness, transparency, and disclosure operate to both a) provide channels of involvement for imprisoned people, inviting them to shape their own data doubles and b) legitimate and galvanize the forms of penal management suggested by the data double as it was constructed via assessment. By being transparent, open, and participatory with one's OASys, justice professionals can attest to greater engagement with prisoners, better accuracy of their assessments and invite imprisoned people into the OASys-suggested framework of their needs, risks and plans for the future. Built in transparency, therefore, legitimates the creation of carceral data structures and their operationalization into forms of penal management.
For imprisoned people, this situation presents few options. One can comply with one's file, faithfully engaging with and following the suite of risk reduction work suggested by the data double produced through an assessment. Or, one can outright resist the file's framework, disputing its contents openly, refusing to engage in risk reduction work and behaving and speaking in ways that defy its rehabilitative pathways, which often leads to intensified punishment or problematization. Lastly, one can seek to dispute the file's contents through procedural channels. This last method deserves further attention, as it highlights the modes of transparency the prison is designed to accommodate.
For one, procedural channels exist to dispute or correct files. These would require involvement of legalistic tactics for the disclosure and correction of data, either involving a legal advocate or engaging an imprisoned person in legal disputation. One might also invite an independent forensic practitioner to conduct a differing OASys, or other assessment, to dispute the characterization of an individual given in a prison-completed assessment. Here transparency-as-accountability draws the individual into the bureaucratic logic of redress. Evidently, this path suffers from the individualized and procedural weaknesses inherent in legislated forms of personal data access and correction. More than this, though, such efforts necessarily appeal to expertise -be it judicial or psychological -to have one's disputes or corrections formally registered. In this form, too, even efforts to "resist" one's file draw an individual into participation with its logic, on fixed terrain. Individuals are left disputing the accuracy or meaning of elements recorded through the OASys, the minutiae of their data double.
The political mobilization of this data double, the inbuilt assumptions of its construction, its managerial imperatives, foregrounding of particular elements of one's life, construction of "risk" and "need" -remain untouched.
This problem extends beyond the semantic. Such assumptions are continually being mobilized to manage, coerce, and condition an imprisoned person. The data double is active in constituting the possibilities of one's everyday life. Indeed, it may play a role in framing how efforts to dispute its very contents might be interpreted. Should one have been characterized as resistant to authority, litigious or manipulative through an OASys itself, the very act of challenging an assessment may be interpreted as evidence of criminogenic risk at work. Even in perfectly transparent conditions, then, the continually active practicability of OASys, and the threat of the next OASys' potential estimation of one's risk, looms large. The data double is continually active: its political and managerial functionality constantly in play, trapping individuals into the logic generated by such forms of assessment.
Crucially, power remains paramount. As Birchall (2016, 2) argues, "data sharing therefore apportions responsibility without power. It watches citizens watching the state, delimiting the ways in which citizens can engage with that data and, therefore, the scope of the political per se." When discussing "open" and "closed" data, Birchall (2016, 6) highlights how "open government data is not excluded from this veillant assemblage" as even these attempts at state transparency operate within an ocular logic of viewing and "sharing of data includes a call to watch and act upon that data." She argues that "both closed and open data contribute to the construction of an anti-politicized data subject and public" (Birchall 2016, 6).
Perhaps this bears relevance no more so than in prison, where channels of redress draw us into heavily depoliticizing ground. Does one comply with the file and follow the path of least resistance? Does one openly resist, to be met with potentially intensifying forms of suspicion, management, and punishment? Or does one dispute data on its own terrain, challenging the accuracy or meaning of individualized records and thus yielding to the OASys' as valid, of operating as a legitimate means of detecting risk and generating plans for its management and reduction?

Conclusion
This article mapped the role of risk assessments in prisons, in order to chart the affordances and limitations of transparency-as-accountability in challenging the power dynamics of carceral data collection. As such, we have demonstrated how transparency is itself incorporated into legitimating prison data collection and penal management, and thus foreclosing attempts to resist one's data double. Using OASys as a case study, we argued that such data-gathering structures mobilize transparency and accountability to legitimate their functioning, bring imprisoned people into the administration of their own penal management, and reproduce carceral imperatives more broadly. Such risk assessments produce data subjects enmeshed in punitive structures of administration and management. At scale, this process operates to foreclose collective attempts to challenge the prison's data structures, and ensuing management strategies, in ways that neatly dovetail with the imperatives of the prison estate as a whole. Data subjects produced through such sociotechnical assessments are not apolitical or neutral, but rather are tied directly to issues of power and administration. They are implemented directly, playing a decisive role at crucial moments within prison governance and decision-making. For us, the issue is not whether the data subjects produced by risk assessments accurately reflect individuals' "risk," "criminogenic need," "offending behavior," etc., but how these categorizations play an active role in managing and conditioning the subjects of assessment. Efforts to dispute, reveal, and correct issues within an individual's assessment are, therefore, restricted in important and significant ways.
At best, transparency and accountability structures can lead to legal disputation over a specific instance of informational accuracy, and decide whether or not there is a significant "mischaracterization" of some aspect of the individual's risk. The forms of redress available through transparencyas-accountability, here, tend to be organized around narrow procedural and interpretive questions, as opposed to substantive structural concerns. The central issue is not the accuracy or inaccuracy of the data -which transparency would be well placed to resolve -but the underlying political and control functionality of the data subject produced, which such procedures of accountability uphold. In prison, a perfectly "accurate" and "transparent" version of the data subject produced through assessment would correspond closely to securitarian imperatives. If, as we have argued, official transparency-as-accountability has its limits -in that the forms of procedural and legal recourse and oversight enacted by transparency are structured by the disciplinary logics of the system itself -what other structures of justice might be possible?
To contend with the failures of transparency and accountability structures within the prison, we have argued that we must overcome various common sense ideas about the potential of transparency as a virtue and mode of redress. Rather than thinking of transparency-as-accountability as a neutral framework for holding systems to account, we must understand these procedures as an expression of the underlying imperatives of the prison. Such analysis may bear fruit more widely than prison itself. Instead, we must ask, what demands in this context might "reduce the power of an oppressive system while illuminating the system's inability to solve the crises it creates" (Berger, Mariame, and Stein 2017).
The efforts of interested scholars on the question of data transparency may want to reckon with this relationship. If transparency can operate to legitimate forms of management, control and punitive bureaucracy, its value as a tactic in resisting institutional injustice is, at the very least, a tempered one. Questions around the forms of control produced by carceral assessments, as well as other forms of data collection, must reckon with broader understandings of that data's use. Embedding understanding of this process in an appreciation of the logics that undergird its function might open up conversations about other potentials for systemic transformation.