I Come before You a Changed Man: “Insight,” Compliance, and Refurbishing Penal Practice in California

Abstract

In 2008 the California Supreme Court forced the state’s parole board to change how it justifies the decision to keep a person in prison. This article combines computational and interpretive analysis of 9,842 hearing transcripts to show how, to achieve compliance with the court, parole board commissioners refurbished an old rehabilitative way of talking about incarceration found in a set of secondary hearing procedures and placed it at the center of decisions. The article considers the consequences of this shift for inequality in incarceration in California. More broadly, it makes the case for focusing on the relationship between administrators and those who can directly intervene in their practices in the name of compliance to understand bureaucratic penal change.

INTRODUCTION

On August 21, 2008, the California Supreme Court issued a pair of rulings that would deeply impact how the state’s Board of Parole Hearings (BPH) justifies its decisions to keep a person in prison. The first ruling, In re Lawrence (2008), invalidated the BPH’s primary rationale for denying parole: the heinousness of a person’s crime. Instead, the BPH had to provide “some evidence” of a person’s current dangerousness to deny parole. But the other ruling, In re Shaputis (2008), held that the facts of the crime could still bear on the board’s task. A person’s current mindset and attitude toward their crime—for example, whether they took full responsibility, showed remorse, or had insight into the causes of the crime—could meet the requirement for “some evidence” of a person’s continuing danger.

Within the boundaries laid out by the court rulings, the BPH could have achieved compliance in many ways. For example, they could have doubled down on explicitly retributive justifications, such as insufficient remorse, for keeping someone incarcerated. They could have turned to actuarial methods aimed at incapacitation, hanging decisions on quantifiable measures like a person’s age and in-prison disciplinary record. They might have adapted emerging restorative justice principles and practices. Or they could have radically reoriented hearings around establishing postrelease supports under a presumption of granting parole. Instead, commissioners began justifying decisions based on whether they thought a person demonstrated “insight” into the causative factors and character defects that led them to commit their crime.

Today, “insight” is at the heart of parole board hearings in California (Hempel 2010; Wattley 2013; Shammas 2019). As this article will show, the shift to “insight” was surprisingly fast for a torpid organization beset by backlogs and staffing shortages. The recent rise of “insight” is even more remarkable when compared against dominant theories of how mass incarceration has shaped bureaucratic penal practices. Scholars argue that since the 1970s, the hyper-incarceration of primarily young, poor, Black and Latino men in the United States reoriented bureaucratic penal practice toward risk and retribution in the face of economic restructuring, White racial resentment, swelling prison populations, and vanishing organizational resources (Feeley and Simon 1992; Beckett 1997; Garland 2001; Harcourt 2007; Wacquant 2009; Hannah-Moffatt 2013; Hinton 2016). These developments replaced mid-century practices aimed at rehabilitating offenders, which, as we will see, fit as a hand to the glove of “insight” as a principle of penal practice.

At the time of the court rulings, California remained on the leading edge of tough-on-crime politics, while the state’s severely overcrowded prisons meant the BPH faced deep organizational strains. In that context, prevailing theories would expect a deeper entrenchment of risk and retribution. How, then, did “insight” so rapidly move to the center of this major penal institution in the twenty-first century, and what are the consequences of this shift?

To answer these questions, this article examines 9,842 BPH hearings from January 2007 to March 2010. Key to the answer will be legacy practices that have persisted on the bureaucratic fringes. Combining new computational text analysis methods with interpretive analysis, I find that “insight” circulated prior to the court rulings in supplemental psychological risk assessments that commissioners reviewed during hearings. In these assessments, clinical psychologists discussed an incarcerated person’s “insight” as a supporting but secondary part of an overall, nonbinding professional determination about that person’s risk to public safety. This use echoed back to the mid-twentieth century when penal administrators discussed “insight” in the context of evaluating rehabilitative treatment. After the 2008 rulings, commissioners moved “insight” from the assessments into the center of the hearings. However, now commissioners increasingly assumed the authority to determine whether a person had insight based on their own judgment instead of relying on the psychologist.

While the term loosened from the psychological assessments, it became increasingly tied to decisions to keep people in prison, constituting a new vocabulary of denial. As a result, to gain release from prison incarcerated people now had to convince commissioners of their insight in the hearing. In practice this meant providing an account of why their crime happened that squared with commissioners’ commonsense assumptions about why someone like them would generally commit a crime. This created opportunities for commissioners to graft their assumptions about criminality and personhood, rooted in a person’s race, class, and gender, onto the language of “insight” in a way that simultaneously erased broader contexts of violence and incarceration. Given that the BPH decides whom to release among the second-largest group of life-sentenced prisoners in the world (Zyl Smit and Appleton 2019, 89), the emergence of “insight” would have large consequences for the practice of incarceration in California.

In showing how an old language of punishment became new again, the article supports and extends recent research arguing that the turn toward retribution and risk has only been partially realized in practice (Lynch 2000; Goodman 2012; Werth 2013; Goodman, Page, and Phelps 2017). Recent work in this vein explains the longevity and resurfacing of old practices through the lens of political struggle or legitimacy (for example, Goodman, Page, and Phelps 2017; Rubin and Reiter 2018; Koehler 2019; Page, Phelps, and Goodman 2019). This article helps to specify the contours of such struggles by emphasizing the role of compliance (Edelman 1992; Mahoney and Thelen 2010; Lara-Millán 2021) in resurfacing old ideas about punishment. Residual routines, lingering on the edges of administrative processes, can keep old and outmoded ideas alive in the routine functioning of bureaucratic organizations. Such accumulated routines allow a bureaucracy that looks unwieldy and inefficient at any given moment to achieve a nimble response to changing requirements of compliance by simply refurbishing existing routines. This explains how an old rehabilitative ideal rapidly reemerged in the foreground of penal practice at the height of mass incarceration in California.

The article proceeds as follows. I begin by developing the theoretical framework within dominant understandings of penal change. I then trace the history of “insight” during the twentieth century before providing legal and organizational context for the BPH at the beginning of the twenty-first century. From there, I introduce the data and analytic strategy. Analyzing word frequencies, I provide evidence that “insight” rapidly moved to the center of hearing decisions as a new “vocabulary of denial” in the six months after the court rulings. Then, drawing on forty case studies of individuals appearing before the board in the study period, I dig deeper into the term’s routine use and practical consequence. The case studies provide evidence that the basis for determining a person’s “insight” migrated from clinical psychologists to commissioners. I buttress this analysis with a word embedding model to show that the term retained the same meaning and conceptual structure through this shift, while it displaced discussion about both the crime’s heinousness and the psychological assessments. I conclude by closely analyzing two hearings to illustrate how commissioners, now determining “insight” themselves, could fit the concept to their assumptions about criminality based on an incarcerated person’s race, class, and gender.

THEORETICAL ORIENTATION

Who should be in prison? This political question has gained urgency in the face of the staggering size and racial disparities of the US prison population (National Research Council 2014). It is also a practical question facing administrators across the penal system who must provide answers on a day-to-day, person-by-person basis. These administrative actors, stretching from the front end of the criminal-legal system with police officers and district attorneys to the back end with clemency panels and parole board commissioners, sit at the intersection between mass incarceration and any individual’s imprisonment. How do these actors answer the question of who should be in prison, and how have their answers changed over time?

Dominant Theories of Penal Change: Risk and Retribution

In the middle of the twentieth century, punishment in America was defined by what David Garland terms “penal welfarism,” or the effort to reform lawbreakers into law-abiding citizens through a mix of punitive sanctions and rehabilitative treatment (2001, chap. 2). Indeterminate sentencing, where a person was to be confined to prison until administrators determined that their rehabilitation was complete, sat as a lynchpin of this approach. At least in aspiration, decisions about whom to incarcerate evaluated the need for individually tailored vocational training and psychological treatment.

By the 1970s, rehabilitative approaches to punishment faced increasing skepticism. Since then, scholars identify two complementary directions toward which penal practices have moved: ensuring retribution and managing risk. First, penal actors became more symbolically punitive in step with changing attitudes toward crime, driven by White racial animus and economic restructuring (Beckett 1997; Garland 2001; Wacquant 2009; Hinton 2016). At the level of bureaucratic practice, these political pressures animated a more vengeful spirit that foregrounded incapacitation and retribution. Second, penal institutions have come to prioritize aggregated risk management practices (Feeley and Simon 1992; Harcourt 2007; O’Malley 2010; Hannah-Moffatt 2013). Shrinking resources and rapidly increasing prison populations drove administrators to latch onto new statistical techniques for classifying and managing populations based on risk. Putting both together, an image emerges of a penal system thoroughly remade by strained organizational resources and increasingly acrimonious and racially charged penal politics.

California in the early twenty-first century seemed to exemplify both trends. On the political front, the state’s penal politics were still in the enduring hold of law enforcement unions and victim’s rights advocates (Page 2011), with an electorate primed to support extreme punishment for violent crime. Indeed, in 2008, voters decisively passed an initiative to give victims more influence in judicial processes (including parole proceedings) and harshen punishment for those convicted of violent offenses, while the same voters would uphold the death penalty in 2012 and 2016. Meanwhile, administrative conditions would suggest a further shift toward risk-based approaches. In 2008, the California prison system was facing a crisis of overcrowding as it hovered near 200 percent capacity. This extreme overcrowding translated into a backlog of thousands of parole hearings, exacerbated by staffing shortages (Legislative Analyst’s Office 2006). Canonical theories of mass incarceration suggest that these conditions would lead to further emphasis on retribution and reliance on aggregated risk management.

However, changes to parole release in California cut against both trends. Scholars have identified the rise of “insight” in parole board decisions beginning around 2008 (Hempel 2010; Wattley 2013). With “insight,” the parole board asks those before them to provide an account of why they committed their crime, to be developed through self-reflection and aided by rehabilitative programming. Canonical theories of penal change fail to explain the emergence of this rehabilitative notion of “insight” at the parole board in 2008.

An Alternative Approach: The Bureaucratic Politics of Compliance

The appearance of “insight” appears surprising partly because the risk and retribution accounts characterize change as a sharp transition—from penal welfarism to a culture of control, from Fordist-Keynesianism to neoliberal governance, from old penology to new penology—determined by seismic broader economic, cultural, and social shifts (Goodman, Page, and Phelps 2017, 5–8). The before and after appear logically distinct and practically incompatible. A growing number of voices question the depth and mutually exclusive nature of these changes. In some parts of the system, penal administrators continue to keep alive rehabilitative ideas and practices that have fallen out of favor in other parts (Lynch 2000; Goodman 2012; Werth 2013). Moreover, the rhetoric of administrators can diverge in dramatic ways from the actual practices of punishment (Lynch 2000; Phelps 2011; Goodman, Page, and Phelps 2017). This is to say that principles and practices of rehabilitation never fully disappeared in the era of mass incarceration.

To explain the perseverance of old ideas over time and the uneven changes to penal practice across place, a new body of work focuses on political contestation and legitimacy (Campbell and Schoenfeld 2013; Goodman, Page, and Phelps 2017; Rubin and Reiter 2018; Koehler 2019; Page, Phelps, and Goodman 2019). This new approach maintains that actors are motivated by relatively fixed political commitments to certain regimes of punishment. They then strategize how best to impose their preferences over others with competing aims amid shifting political coalitions and broader social conditions. Success hinges on legitimacy and political capacity, or what ideas and actors are taken seriously at any given point in time.

While helpful for understanding legislative and popular debates over the nature of punishment, this approach struggles to explain bureaucratic change specifically. First, legitimacy and political capacity appear as latent and abstract qualities of individuals and groups, making it difficult to analytically separate them from outcomes. This is particularly true in administrative arenas lacking open, deliberative debate. Second, there are no clear criteria for determining the relevant set of actors. And third, there is no concrete expectation for when or how past practices influence the present.

To clarify the bureaucratic politics of penal change, I argue for greater attention to compliance. Maintaining compliance with shifting external directives is an important driver of organizational change (Edelman 1992; Mahoney and Thelen 2010, 10–14; Lara-Millán 2021). Building on the literature on political contestation and legitimacy, I suggest that focusing on compliance helps to clarify the dynamics of bureaucratic change in penal institutions on three counts. First, it provides identifiable criteria for determining when organizations are in or out of compliance, as opposed to trying to determine levels of legitimacy or political capacity. Second, it focuses attention on the specific actors who can formally declare an organization out of compliance and directly intervene in its practices. This can refer to a wide range of actors, from courts to oversight committees to legislators to funders. But the criterion for inclusion depends on whether the actor has a formal relationship of compliance with the organization. Third, it focuses on the routine practices of actors rather than the abstract principles they may profess. Compliance concerns how administrators carry out the imperative to keep things running amid shifting requirements. Ultimately, the consequences of broader political struggle for bureaucratic practice must be mediated through the dynamics of compliance.

Such an approach adds new depth to the significance of enduring legacy practices amid penal change. Rubin (2016) argues that new practices “displace” rather than eliminate the practices over time in prisons, leading to a process of “penal layering” (422). This article specifies a set of conditions for how such displaced practices can gain newfound relevance. Meanwhile, scholars of legal endogeneity stress how organizations themselves can define legal compliance in the interest of the organization (Edelman 1992; Edelman, Uggen, and Erlanger 1999; Dobbin and Kelly 2007; Edelman et al. 2011). This work focuses on the emergence of new practices to symbolically signal compliance to outside actors. As I will show, the BPH did not create a set of new routines to signal compliance. Rather, they refurbished an older practice from the grab bag of procedures accrued over decades. The article shows how psychological evaluations conducted for each parole board hearing literally put “insight” into the mouths of the commissioners, making it readily available as an alternative rationale for punishment as the requirements of compliance shifted. This unwieldy set of practices, within an organization facing significant backlogs and staffing shortages, provided the basis for a remarkably nimble response to shifting requirements of compliance. In the next section, I consider the question: what, exactly, was this ready-made conception of punishment?

HISTORICAL AND PRESENT CONTEXT

A Brief History of “Insight” as Penal Welfarism

While largely undocumented by contemporary scholars, penal administrators drew on “insight” as a governing rationality throughout the middle of the twentieth century. For example, the Second National Conference on Parole in 1956 brought together penal experts from across the country to discuss the leading edge of correctional practice. In summarizing best practices for determining whether to release a person from prison, the conference proceedings stated, “Growth in insight into his own problems and into the real motives for his offense should be expected” (Parole in Principle and Practice 1957, 107). Meanwhile, a mid-century empirical study of parole board decisions in three states found that “[t]he indication of parole success most frequently searched for at parole hearings in Michigan and Wisconsin is evidence of a change in the inmate’s attitudes toward himself and his offense. This is commonly referred to as an inmate’s gaining ‘insight’ into the problem which caused his incarceration” (Dawson 1966, 250). Administrators also invoked the idea in their public statements. As the chair of New York’s Board of Parole explained in a 1970 essay in the journal Federal Probation, “It is not unusual for a person, who appears on paper to constitute a good parole risk, to talk himself out of parole by venting his inner attitudes or his total lack of personal insight as to his problems under questioning by Board members” (Oswald 1970, 28).

Likewise, “insight” infused penal practice in California. Walter Gordon, the first African American man to graduate from Berkeley’s law school and head of California’s parole board from 1946 to 1955, wrote that the goal of providing people with “insight” motivated the state’s prison programs: “The treatment facilities within the institutions consist of educational and vocational opportunities, good and adequate work programs, psychiatric therapy, and specialized professional services. The attempt is made to give the inmate insight into his personality problems and methods of overcoming them” (Gordon 1947, 217). If taken on its own terms, “insight” suggested a deeper aspiration of this earlier era: to effect a deeper change in the incarcerated person, only superficially connected to the commitment offense. However, Gordon explained that the parole board could not simply rely on a person’s program record to determine whether they had developed insight. The board had to evaluate the success of this treatment for themselves:

A treatment program will not be effective unless the inmate is willing to accept the opportunities offered. It is in questioning him when he appears before the Authority that we endeavor to gain an insight into his thinking. His thinking and acting should determine to what extent he has accepted the treatment program. Taking into consideration all of these factors, we try to fit the punishment to the man and not the crime. (218, italics in original)

The parole board would need their own “insight” into a person. Thus, regardless of the facts of the crime and the person’s in-prison record, how they appeared “on paper”—the success of this treatment, and hence whether a person should be released—would rest on the potential parolee’s ability to convince commissioners of their new thinking in a parole hearing.

“Insight” formed part of a larger parole system that had come under intense criticism by the 1970s as arbitrary and discriminatory in process and outcome (Garland 2001). Over the ensuing decades “insight” would recede to the background of California’s penal system. Since the 1970s, Aviram (2020) writes, “at every junction, the California legal process, heavily shaped by the nascent victims’ rights movement and the legislative initiative process, shifted away from a logic of professional, clinical assessment of rehabilitation toward a deeply politicized process largely reliant on the manipulation of public emotions and fear” (12). ¹ As a result, parole hearings by the turn of the twenty-first century centered increasingly on the facts of a person’s crime. The next section considers what the BPH looked like at the tail end of this long process of transformation.

Insight and the California Board of Parole Hearings Today

The BPH determines who should get out of prison for all people serving indeterminate life sentences in California—approximately thirty-five thousand people, or a little over a quarter of the state’s prison population (Nellis 2017, 10). Most “lifers” in California are serving indeterminate sentences for violent crimes such as murder, manslaughter, rape, and kidnapping (Nellis 2017). Indeterminate life sentences have a fixed minimum term and unbounded maximum term, such as “twenty-five-to-life.” When a person reaches their minimum term (e.g., twenty-five years), they become eligible for parole. Meanwhile, the commissioners who will decide whether to parole that person exercise wide discretion in applying statutory parole suitability factors. They tend to come from prior careers in the criminal-legal system, such as corrections and law enforcement. In hearings, BPH commissioners must make sense of the complexity and ambiguity of a person’s crime and life to reach an unambiguous, high-stakes decision about whether to release that person from prison.

The parole board has two important formal compliance relationships: the governor and the courts. The governor holds veto power over all BPH decisions. In the years leading up to 2008, California voters elected idiosyncratic Republican Arnold Schwarzenegger as governor. Surprising political observers at the time, Schwarzenegger embraced a rhetoric of rehabilitation (Warren 2005). Yet he also maintained strong alliances with victims’ rights groups, and it is doubtful that his professed commitment to rehabilitation dramatically impacted administrative practices inside prisons. Certainly, his public embrace of rehabilitation starting in 2005 did not immediately translate into parole board decisions, which continued to rest heavily on the “heinousness” of a person’s crime in his first years.

Instead, the power of judicial review would prove more decisive. On August 21, 2008, the California Supreme Court held in Lawrence that the BPH had to make decisions based on a person’s current dangerousness, not on historical factors that could never change. This was a direct rebuke of parole denials that rested on only the heinousness of a person’s crime. Released on the same day, Shaputis ruled that parole decisions could rely on a person’s current mindset and attitude toward the crime, which included things like whether the person showed remorse, took responsibility, or demonstrated insight into the crime. If the BPH did not comply, the courts would begin reversing individual parole decisions. To achieve compliance, then, The BPH faced a choice. Within the framework of Shaputis they could have doubled down on remorse, or moved to actuarial decision-making approaches, or gone further afield to institute restorative justice principles. Or with Lawrence the door was open to make more radical changes, such as reorienting parole hearings toward establishing postrelease supports for people under the presumption that they would be granted parole.

Instead, following the rulings, BPH decisions came to hinge on whether a potential parolee possessed “insight” into their crime (Hempel 2010; Wattley 2013; Shammas 2019). That is, commissioners began to expect those before them to articulate knowledge of the causes of their crimes to gain release. Scholars argue that the “rehearsed authenticity” (Aviram 2020, 105) that commissioners expect in discussions of insight varies along racial, class, and gender lines (see also Shammas 2019; Greene and Dalke 2021).

The number of people eligible for release through the parole board has expanded considerably since 2008. The state came under pressure to decrease the prison population following the 2011 US Supreme Court decision in Brown v. Plata, and people sentenced under the state’s three strikes law, passed in 1994, have become eligible for parole hearings. Meanwhile, the board has undergone changes in political and administrative leadership, seen serious attempts to diversify board composition and provide more professional training, and adopted more formalized hearing procedures. It is also worth noting that the rate of parole grants began to increase noticeably during the study period, rising from 1.9 percent of scheduled hearings in 2007 to 8.7 percent in 2009 and continuing to rise to 16.3 percent in 2021. ² It is difficult to attribute this change to any single cause given the multiple political, organizational, and legal changes, but undoubtedly more people have gained release through the parole process after the court rulings.

One thing that has not changed is the centrality of “insight” to parole board decisions. As a legal guide by a prominent legal nonprofit representing people up for parole states in its 2022 parole preparation guide:

It is very important that the person [up for parole] be able to demonstrate that they have gained a clear understanding of their background prior to the life crime (including family relationships and prior criminal or juvenile record), the circumstances leading to the crime, how they have resolved and can prevent a relapse to the circumstances that led them to violence…. A person’s ability to understand and discuss these factors determines whether or not the Board finds that they lack “insight.” If the parole applicant does not understand these factors, they will be denied parole, no matter how much time they have served and no matter how spotless their disciplinary record is. (UnCommon Law 2022, 5)

Many people in California prisons spend significant time developing accounts of their insight for their parole board hearings. This can involve tremendous effort and a process of deep personal transformation, while the uncertainty and arbitrariness of gaining release can inflict a deep psychic toll (Aviram 2020). Today many state prison rehabilitation programs aim to help incarcerated people develop insight in preparation for the board, though many people still receive woefully insufficient support (Petek 2023, 12–13). As more people have gained release through the parole board, their path to release has required them to convince commissioners of their “insight.” Meanwhile, many remain incarcerated for simply being unable to do so.

No peer-reviewed research has assessed whether the board’s conceptualization of “insight” corresponds with any specific measure of future dangerousness, and the broader evidentiary base is thin. Among the general literature on recidivism, one line of research, building on Maruna (2001), relates reincarceration to whether a person holds a “redemption” self-narrative. It is worth noting that such narratives can contain an extraordinary range of content and involve a complex relationship with social context (Maruna and Liem 2021). Further, as a predictor of “dangerousness,” it is difficult to adjudicate between whether a person’s self-narrative shapes a person’s risk, or simply provides a script for justifying or rationalizing actions after the fact, or imperfectly corresponds to a different underlying aspect of risk (Sampson and Laub 2016). In any event, it is unclear how closely parole board commissioners’ conception of “insight” corresponds to this research, or how one would clearly and consistently demarcate between a “good” self-narrative and a “bad” self-narrative in a charged setting with such large consequences attached. Put another way, no validated procedure exists to consistently identify a person’s “insight” in relation to their risk of committing future harms.

Given how central “insight” now is to parole decisions, Wattley (2013) points out one more curiosity: “It bears noting that the term ‘insight’ never actually appears in the Board’s guidelines for determining parole suitability” (273). In short, the new centrality of “insight” was far from fated. How, then, did it become so central to parole release decisions at the height of mass incarceration in the state? To answer, I turn to the hearings themselves.

DATA AND METHODS

In the following analysis I rely on transcripts from all 9,842 BPH hearings that resulted in a grant or denial in the seventeen-month window around the August 2008 rulings, ranging from January 2007 to April 2010. ³ Each transcript contains the dialogue that occurs on the record between the panel (consisting of two BPH commissioners), the person up for parole, and other hearing participants (such as lawyers or victims’ next of kin). Following an off-the-record deliberation period, each transcript captures the commissioners’ decision and their explanation for the decision. The transcripts are the official legal record for the hearing. They are read and referenced by the BPH and the governor’s office in reviewing decisions, by commissioners in future hearings, by judges in legal cases, and by the incarcerated person to prepare for future hearings. In short, they are key documents for observing the dynamics of compliance.

The use of the transcripts comes with some caveats. First, there is no systematic information on the race of the person up for parole, precluding analysis of race across all decisions (though a person’s race is often identifiable through close readings). Second, the transcripts do not capture the embodied interactional dynamics of a hearing, such as tone, body posture, and speech delivery. These are all important interpretive cues that contribute to the meaning of commissioners’ words. However, relying on the transcripts mirrors the perspective of actors evaluating compliance after the fact. Third, the transcripts provide no way to tell whether in any instance the commissioners are employing motivated reasoning—justifying a decision reached on separate grounds that are not articulated in the hearing—or attempting to sincerely reconstruct how they arrived at a decision. Fourth, this analysis cannot determine whether the rise of “insight” resulted in more people, or different people, gaining release. Instead, the focus is on how the justification for parole decisions changed over this period.

To analyze commissioners’ decision-making justifications, I take a “computational grounded theory” approach that situates computational analysis within and alongside traditional interpretive techniques (Nelson 2017). I aim to take the best of both: pairing pattern detection at formerly unimaginable scale with deeper nuance that comes from close reading. I deploy a series of computational text analyses based on word frequencies and word embedding models that examine all of the transcripts. I also use traditional interpretive techniques to analyze forty case studies of individuals who appeared before the board during my study period. Because of the range of methods involved, I will introduce each as it arises. I additionally provide a comprehensive overview of the computational techniques in Online Appendix A.

THE EMERGENCE OF INSIGHT

The first question to address is whether the Lawrence and Shaputis rulings did, in fact, have an impact on how often “insight” came up in parole decisions. In this section I examine word frequencies to show how talk about “insight” became more frequent across and within parole decisions in the six months following the rulings, particularly in parole denials. In this, I demonstrate that “insight” rapidly became a new vocabulary of denial following the rulings.

First, looking across hearings, Figure 1 shows the percentage of decisions in each month mentioning “insight.” The figure shows a flat trend in the months leading up to just before the cases were argued before the court in May 2008 (indicated by the gray dashed line). Then the graph shows a strong upswing. ⁴ At that point, the percentage of decisions where “insight” is mentioned rises from 44 percent in June 2008 to 81 percent by January 2009, nearly doubling in a short six-month period. (Online Appendix B.1 presents evidence that this was not driven by any individual commissioner.)

FIGURE 1. Portion of Transcripts Mentioning "Insight" by Month for All Hearings between January 2007 and April 2010.

Following the rulings, the word also appeared more frequently within decisions. Before the rulings, the word appeared an average of 2.03 times in decisions that mentioned “insight.” After the rulings, it appeared on average 3.12 times. This difference in the word’s frequency between the two periods is statistically significant at p = 0.001 on a two-sided Kolmogorov-Smirnov test comparing the cumulative probability functions for mentions in both time periods (statistic = 0.241, p < 0.0001).

Last, the term “insight” changed its association with hearing outcome. Table 1 shows how often the word “insight” appeared in grants and denials before and after the rulings. To test association with outcome I conduct two chi-squared tests that evaluate the relationship between the use of “insight” and hearing outcome before and after the court rulings. Before the rulings, “insight” was not statistically associated with outcome (at p = 0.05). Yet after the rulings, “insight” becomes positively associated with denials (at p = 0.001). Put another way, following the rulings the word increasingly is used in the context of keeping a person in prison.

TABLE 1. Mentions of “Insight” by Hearing Outcome before and after the Court Rulings

For a highly bureaucratized organization, facing severe resource constraints, this is a remarkably nimble transformation. The term “insight” was in use before the Lawrence and Shaputis rulings. Yet within months following the rulings, commissioners began to talk about “insight” in more hearing decisions, and more frequently within decisions, as it became increasingly part of a vocabulary of denial—a term used to describe a person’s unsuitability for parole. Yet this analysis also suggests that “insight” was in circulation prior to the rulings, albeit in a more minimal role. The next section turns to a close reading of transcripts to understand how commissioners used “insight” in decisions before and after the rulings.

Shifting Authority

To understand when and where “insight” came up, I conducted a close reading of transcripts for forty individuals who appeared before the BPH during the study period. I find that the term circulated before the hearings through supplemental assessments produced by psychologists to determine the potential parolee’s risk of reoffending. These assessments linked a person’s future dangerousness to a host of concerns, including a person’s “insight”: if a person could not explain why they did what they did, they could not be trusted to avoid doing it again. Commissioners would read directly from these evaluations during hearings, literally putting “insight” into their mouths. After the rulings, the commissioners picked up this use of the term and its underlying logic, and increasingly began to make their own determinations about a person’s insight based on what happened in the hearings. In the analysis below I lay out what this looks like in the hearings, before returning to computational text analysis to provide further evidence of this change.

A note on close reading

Following a “computational grounded theory” approach (Nelson 2017), I conducted the close reading after initial computational analysis. I followed a stratified sampling strategy to select forty initial transcripts based on three characteristics of the hearings: whether “insight” came up in the decision, whether the person was granted or denied parole, and whether the hearing was held before or after the Lawrence and Shaputis rulings. I randomly sampled five transcripts from each of the eight bins provided by this stratification strategy, resulting in transcripts for forty different incarcerated people. I then followed each of these people across all their hearings during the study period, reading a total of eighty-two transcripts (totaling 7,586 pages). Each individual in the sample appeared before the board between one and four times. These forty case studies allowed me to follow individual trajectories and trace how “insight” took shape over this period.

My analysis followed an iterative coding and memoing process. I first read each transcript holistically to write a descriptive memo for each hearing and collect information on the potential parolee’s profile, including participation in prison programs, disciplinary history, parole plans, and their race, class, and gender. ⁵ Midway through each transcript, after finishing the proceedings portion of the transcript but before reading the decision, I wrote an analysis predicting the hearing outcome and whether “insight” would come up in the decision. This allowed me to assess and deepen my own understanding of how the hearing proceedings related to the decisions. After reading the decision, I wrote another analysis of the rationale behind the decision, how it linked to the earlier proceeding, and how it compared to my prediction. Meanwhile, I also identified each instance of key words such as “insight,” “remorse,” “risk,” and “heinous,” collecting information on the speaker, source, point in the hearing, and relationship to outcome. This provided an initial set of themes and hypotheses that I investigated through further iterative coding and memoing to arrive at the analysis presented here.

Before the rulings—clinical evaluations

The word frequency analysis previously showed that “insight” was in use in a minority of hearings before Lawrence and Shaputis. The close reading reveals that commissioners commonly discussed “insight” prior to the rulings in the context of supplemental psychological risk assessments. Every person up for parole receives an assessment prepared by a clinical psychologist evaluating their overall risk of future violence. In every hearing, commissioners review the assessment and note relevant portions, often reading directly from the write-up.

Not all psychological assessments focused on insight, and not all commissioners focused on “insight” when reviewing the assessments. Yet in hearings commissioners would occasionally read verbatim the psychologist’s explanation for why insight was important alongside the psychologist’s evaluation of a person’s insight. This is the case in the following excerpt from the 2007 initial parole board hearing for Stewart Holden, ⁶ who was serving a fifteen-to-life sentence for manslaughter. During the hearing the deputy commissioner read directly from the psychologist’s assessment:

The doctor closes with Comments and Recommendations: ‘ … Also of concern is his limited insight into the seriousness of his crime. It is of importance that Mr. Holden takes the time to seriously reflect on his past lifestyle so that he gains additional insight into what lead to the consumption of drugs and the perpetration of violent act [sic] against another human being. Without benefit of an insight intense to self-reflection, he is a greater risk for acting in a manner that would lead him to being returned to prison.’

Here the psychologist determined that Mr. Holden does not have insight into his “past lifestyle.” The psychologist argues that self-reflection on the causes of the crime is important for a person to desist from future crime, and hence insists that Mr. Holden’s lack of insight increases his risk of future violence. This rationale echoes the explanations provided by penal practitioners from the middle of the twentieth century presented earlier (if, as noted earlier, it is on thin empirical ground). In Mr. Holden’s hearing, the presiding commissioner would cite this same passage in explaining the panel’s decision to deny parole. This lack of insight supplemented other reasons: that the crime was vicious, that Mr. Holden had a lengthy criminal history, that he had participated minimally in programs, and that he had amassed numerous disciplinary write-ups.

Notably, “insight” played only a secondary role in the psychological assessments. The assessments aimed to provide expert judgment of a potential parolee’s overall proclivity for future violence. The prevailing legal guide of the time describes the assessments as such:

Prior to most parole hearings, either a psychologist or a psychiatrist will prepare a psychological profile of the prisoner, noting the prisoner’s expressions of remorse and any psychological problems related to the offense (or not). There is also a brief discussion of the prisoner’s psychosocial history and family background. Most importantly, the psychological report will include the clinician’s assessment of the prisoner’s danger potential, often in terms of low, moderate or high risk to public safety in comparison to other people in general or other prisoners. (McKay and the Prison Law Office 2008, 237)

“Insight” does not appear in the above description at all. The main purpose of the assessments was to evaluate potential for future crime, and in practice the term factored into assessments to the extent that the psychologist deemed it relevant to this end. ⁷

Since the 1980s, the BPH had minimized the role of psychologists in the hearings and shown increasing attention to the incarcerated person’s statements of remorse (Aviram 2020). But the psychological assessments kept a residue of rehabilitative logics alive in the hearings. Across the case studies, commissioners would cite past assessments dating as far back as the early 1990s that discussed a person’s insight. In a literal sense, the psychological assessments put the vocabulary of “insight” into the mouths of commissioners during this period. As we will see, the California Supreme Court’s rulings would create an opportunity for “insight” to return to the foreground of decision making.

After the rulings—commissioner determinations

After Lawrence and Shaputis, commissioners began questioning and contesting insight in the hearing proceedings, while in the decisions, commissioners increasingly relied on their own judgment to determine whether a potential parolee had insight. To do so, commissioners tied their assessment of insight to a person’s performance in the hearing itself. The underlying rationale of “insight” mirrored the psychological reports. Yet it came from a different source—the commissioners’ judgment—and took on new implications as a vocabulary of denial.

Most basically, after Lawrence and Shaputis commissioners started making explicit references to insight in their questioning. For example, in a 2010 hearing for Jose Gonzalez, who was serving a fifteen-to-life sentence for his involvement in a retaliatory shooting, the presiding commissioner wanted to hear about how Mr. Gonzalez became involved in the crime. This was Mr. Gonzalez’s fifteenth subsequent hearing, and he had previously had a parole grant reversed by the governor, though the board had denied him parole at his most recent prior hearing. The commissioner set up his question like this:

Let me go ahead and frame it so that you understand where I’m going. You’re working. You’re providing for your family. You’re raising children. And other than having some problems with alcohol you didn’t have any adult assaultive behavior. And where I’m going with all of this is to the area of insight into what would’ve allowed you to get involved.

The commissioner bluntly indicates he wants to hear about insight, meaning he wants Mr. Gonzalez to provide an account of why he participated in the crime. This trend is reflected across the eighty-two transcripts I read as part of the forty case studies. While commissioners asked explicitly about insight or directly commented on a person’s insight in the proceedings of only two of the thirty-nine transcripts that came from hearings conducted before the court rulings, they did so in seventeen of forty-three postruling transcripts. That is, commissioners increasingly discussed insight with people up for parole directly in the hearings, beyond the context of the psychological assessments.

In their decisions, commissioners then began placing greater emphasis on their own judgment about an incarcerated person’s insight. We can see this in the following hearing from 2009. Junseo Park was sentenced to life in prison for attempted murder, and this was his second time before the board. Despite his good behavior, participation in self-help programs, and developed parole plans, the panel would deny him release again. In the decision, the presiding commissioner explained,

[Y]ou seem to lack insight into the causative factors which lead all this to you, sir … looking at the Shaputis case, that is vital and very important for a Panel to have an understanding as to the level of insight you have into this crime, as to the level of responsibility, because it is a predictor of your safety in the community, and you need to be able to demonstrate that…. It’s an issue of heart. It’s an issue of what you understand your – what you’ve done and are able to articulate that to a Panel.

The commissioner’s underlying logic shows an affinity with the psychologist’s statement presented in the preceding section: if a person lacks understanding of why he committed his crime, he remains a risk to public safety. Yet here the commissioners reached their own determination. In the hearing Mr. Park could not “articulate” an understanding of the crime—either verbally or emotionally—to the commissioners’ satisfaction. ⁸ That, in turn, provided grounds for denial. The commissioner explicitly signaled the relationship between these grounds for denial and compliance with the Shaputis ruling.

Across cases, commissioners also showed an increasing tendency to make statements about a person’s insight that ran counter to what the psychological assessment found. In five of the twenty-three decisions following the rulings that mentioned insight, commissioners reached a conclusion about “insight” on their own that directly contradicted the psychologist’s determination of insight discussed earlier in the hearing. There were no such cases among the nineteen preruling transcripts that mentioned insight. Whether a person was to be released from prison increasingly hinged on what commissioners made of that person’s insight, making the concept central to release decisions in a way it simply was not prior to the court rulings.

Semantic Continuity and Substantive Change

So far, I have argued that commissioners achieved compliance by simultaneously elevating an existing principle of decision making from a secondary to a primary role while changing the basis of evaluation from the psychologists to the commissioners. In this section, I return to computational text analysis to explore and rule out other possibilities. As an alternative to what I have argued so far, it is possible that the commissioners cynically appropriated “insight” as a new label for their old decision-making rationales. It is also possible that commissioners simply substituted “insight” for a word with similar meaning that was widely used prior to the rulings. I consider both possibilities in turn.

To adjudicate whether the increase in the use of “insight” reflects an appropriation of the word independent of its original meaning, I turn to a historical word embedding model. This will allow me to explore how the discursive context of insight and other words shifted before and after the rulings. If the commissioners began using the word in a novel way, I would expect to find a new discursive context around the word. If they maintained the same meaning before and after the rulings, I would expect the discursive context to remain the same.

Before presenting the analysis, what is a word embedding model? In a word embedding model each unique word in the vocabulary of a corpus is represented as a vector of numbers based on how frequently it co-occurs alongside other “context” words in the text. These vectors provide coordinates for locating (or “embedding”) words in a continuous, multidimensional embedding space. Words that are used in similar ways tend to appear close to each other within this embedding space, by virtue of tending to appear in similar contexts in the corpus. As such, a word embedding model inductively describes relationships between words that may not appear near each other in the corpus but are used in similar ways to talk about similar topics.

For example, given a corpus of daily newspaper stories, we would expect “cloudy” and “sunny” to appear in sentences discussing weather, meaning they would have a substantial overlap in their contexts. As a result, in an embedding model trained on that newspaper corpus, we would expect “sunny” and “cloudy” to be located near each other in embedding space (in other words, we would expect them to be neighbors) relative to other words that might appear in different contexts, like “carrots” or “democracy” or “basketball.” If words are neighbors, this does not mean they have synonymous definitions. Rather, they have overlapping meanings. “Sunny” and “cloudy” are not synonymous, but both are adjectives to describe a certain quality of weather.

If we were to directly count how often words co-occur with context words, our embedding model would grow quite large, because the dimensions equal the number of unique words in the vocabulary. This creates very large and sparsely populated embeddings that pose practical and mathematical problems. To address this, I rely on an unsupervised natural language processing algorithm known as word2vec to create dense, smaller-dimension word embeddings (Mikolov et al. 2013). Word2vec is increasingly popular across the social sciences for creating compact, high-quality word embeddings (for example, Kozlowski, Taddy, and Evans 2019; Lucy et al. 2020; Rodman 2020). The word2vec model is randomly initialized, meaning that it incorporates some inherent randomness. To quantify uncertainty caused by this, I follow a procedure called bootstrapping that generates multiple sets of embeddings trained on resampled variations of the corpus (Antoniak and Mimno 2018). I explain in greater detail the process of creating the embeddings in Online Appendix A.3. I also present an alternate approach in Online Appendix B.2 that is based directly on co-occurrence counts and provides substantively similar results.

By creating word embeddings from texts from sequential historical periods, I can examine how words shift within embedding space over time. Scholars have linked shifts in embedding space to changes in word meaning. For example, historical embedding models have successfully identified the shift in “broadcast” from its use in the 1850s meaning to sow or spread toward the post-1950s use associated with TV and radio, or the change in the word “twilight” after 2009 to refer to the book and movie franchise in addition to the time of day (Kulkarni et al. 2014; Hamilton, Leskovec, and Jurafsky 2016a).

To evaluate whether commissioners began using “insight” in a novel way, or in ways that were substantively in line with the term’s prior use, I generated two sets of word embeddings: one from the decisions before August 21, 2008 (the day the rulings came down), and one from the decisions after. The embeddings in each set are trained only on the portion of the transcript where the commissioners announce and justify their decision, not on the question-and-answer proceedings. I then compare the neighborhood of words in the embedding space around the target word before and after the rulings. Neighborhood change is an indication that a word’s context changed in the text (Hamilton, Leskovec, and Jurafsky 2016b). Because the embedding space is still high-dimensional, I follow the standard practice of using cosine similarity to measure distance between words.

I present results in Table 2, which displays the fifteen closest neighbors to three given target words before and after the court rulings. My ultimate target word of interest will be, to no surprise, “insight.” Before proceeding to “insight,” I present two examples to establish how the models work. First is a case where we would expect the word to have the same meaning before and the after the rulings, and second is a case where we would expect the word to change meanings.

TABLE 2. Word Embedding Neighborhoods before and after the Court Rulings

Note: Unique neighbors appearing in the target word’s neighborhood only before or only after the rulings are in italics. Because fifteen words is an arbitrary cutoff for the neighborhood size, the number of unique neighbors does not exactly align with the magnitude of the target word’s shift in embedding space.

I begin with the word that we would not expect to change meanings: “mother.” Commissioners frequently discuss mothers in decisions, both in the context of family history and as a reliable source of material and emotional support for people in prison. We would not have expected the parole board to redefine the meaning of motherhood over this period. If the way commissioners used “mother” remained consistent before and after the rulings, we would see little change in the word’s neighborhood. Table 2 shows that this is the case. The embedding models suggest that the word remained used in ways consonant with other family members like “father” and “sister” across the time periods.

On the other hand, if commissioners dramatically redefined a word between the two time periods, we would expect the neighbors of that word to change dramatically as well. We can see this in a word we would expect to change over this period: “Lawrence,” the name of one of the court cases that effected the emergence of “insight.” As Table 2 illustrates, before the rulings “Lawrence” was used in ways similar to other proper names like “Leroy” and “Werner.” After the rulings, however, “Lawrence” moves into a neighborhood of specific legal cases (e.g., its companion case “Shaputis,” or important earlier cases such as the “Dannenberg” case [In re Dannenberg (2005)] or the “Rosenkrantz” case [In re Rosenkrantz (2002)]), general legal terminology (e.g., “supreme” and “court”), and legal terminology specific to the case (e.g., “nexus” and “loses,” as in “the commitment offense loses predictive value over time as a nexus to current dangerousness”). This new neighborhood captures the new legal meaning “Lawrence” took on after the ruling.

Where does “insight” fall on this spectrum? As Table 2 shows, the neighborhood of “insight” changed little before and after the rulings. While the relative order of the words moved around (for example, “remorse” drops from being second-nearest to eighth-nearest), there is substantial overlap between the two sets of neighbors. The neighbors refer to a constellation of self-knowledge (for example, “understanding,” “underlying,” and “causes”), emotion (“remorse” and “empathy”), and moral culpability (“minimization” and “creditability”). Qualitatively, it appears that the word “insight” largely stayed put in embedding space before and after the rulings. That is, although “insight” moved to the center of decisions and into the domain of the commissioners after the rulings, the semantic use of the term remained similar to how the psychologists used it before the rulings.

This qualitative conclusion holds up under more formal specification. Using a technique developed by Hamilton, Leskovec, and Jurafsky (2016b), I measure the distance from each target word to its fifteen closest neighbors within each period. I then compare the cosine similarity scores across time periods to create a cosine similarity score of neighborhood change. ⁹ A score of 1 indicates that a word did not shift in embedding space relative to its neighbors. The closer to 0 the score gets, the more a word moved relative to its nearest neighbors (see Online Appendix A for further detail). Figure 2 presents the results. For “mother” the cosine similarity between time periods is 0.996, reflecting that the term stayed put. Meanwhile for “Lawrence” the cosine similarity is 0.849, reflecting that it moved neighborhoods and took on a new meaning. For “insight” the cosine similarity is 0.979, placing it much closer to “mother” on this spectrum. For further reference, the average cosine similarity of the thousand most common words is 0.957. This suggests that “insight” changed little in semantic meaning between the two time periods.

FIGURE 2. Neighborhood Change Score before and after the Court Rulings.

Both qualitatively and quantitatively, the word embedding analysis provides evidence that the semantic use of “insight” before and after the rulings remained quite stable. At least discursively, commissioners used the term in the same way after the court rulings as they had before the rulings. In other words, the models provide evidence that “insight” was not a vacant term the commissioners began invoking in novel or discursively arbitrary ways, but an extant idea about punishment that moved closer to the heart of the hearings.

The next possibility to investigate is whether commissioners substituted “insight” for another similar or equivalent concept in widespread use prior to the court rulings. That is, commissioners could have been talking substantively about “insight” before the rulings using a different word with a similar meaning (or vice versa). To consider this possibility, I look at how the frequency of use for the whole neighborhood of insight changed over the study period (and I provide further analysis in Online Appendix B.3).

Figure 3 shows the average percentage of sentences in decisions by month that contained “insight” or one of its closest fifteen neighbors from the “before” period (the same neighborhood presented in Table 2). If “insight” represented a slight tweak within a larger conceptual approach to the hearings that remained unchanged, I would expect there to be no changes around the time of the rulings. Instead, the percentage of sentences in decisions mentioning “insight” or other words in its neighborhood doubles around the time of the court rulings. For context, Figure 3 includes two more word neighborhoods. First is the neighborhood around “heinous,” which the rulings forbid commissioners to use as the only justification to deny parole. With “heinous,” the figure reveals a noticeable inverse pattern compared to “insight.” As discussion of “insight” and related terms increased, discussion about the heinousness of the crime decreased. Second is the neighborhood around “assessment,” referring to the psychological assessments. ¹⁰ In 2007, the trends for “insight” and “assessment” track each other, with a slight but noticeable rise in the discussion of “assessment” in the months prior to the rulings. However, after the rulings this trend reverses: commissioners begin to talk more about the neighborhood of terms around “insight” while discussion about “assessment” and related words tapers off. This is consistent with “insight” loosening from discussion of the risk assessments in decisions postrulings.

FIGURE 3. Percentage of Sentences Mentioning Neighborhood of Terms (15 nearest neighbors before court rulings).

To summarize, the word embedding model provides further evidence that “insight” became a new way to justify parole decisions rooted in existing practices. The term circulated before the Lawrence and Shaputis rulings in supplementary psychological assessments. To comply with the rulings, commissioners inserted it into the center of parole determinations. While maintaining the earlier language and logic, the commissioners loosened the term from the assessments to increasingly make their own judgment about a potential parolee’s insight. This represented not a hollow appropriation, nor a slight tinkering with existing decision rationales, but a more fundamental reconfiguration of the board’s decisions. In the final section I turn to the question: what are the consequences of this shift in practice?

In Search of Insight

I now present two hearings to illustrate insight in action after the Lawrence and Shaputis rulings. Examining individual hearings highlights how the elements of “insight” came together holistically in practice in ways that were consistent across cases. After Lawrence and Shaputis, the commissioners came to expect people to provide a causal account of the crime that fit with the commissioners’ common sense. Commissioners overwhelmingly came from corrections and law enforcement backgrounds, deeply shaping assumptions about who commits crime and why. This section suggests how “insight” may have opened the door for commissioners to apply that common sense in ways that relied on assumptions rooted in people’s race, class, and gender, while simultaneously obscuring the role of social context.

Lacking insight

The stakes were most stark where the person claimed innocence, where there was ambiguity surrounding what happened, or where the person simply could not remember the crime. David Alarcon falls into this last category.

One fall night Mr. Alarcon was heavily drinking in the parking lot of a convenience store in California’s Central Valley. As late evening turned to early morning he walked into the store and stabbed the attendant multiple times. Twenty years later, well into his seven-to-life sentence for attempted murder and after a string of parole denials, Mr. Alarcon was doing what was expected of him: he hadn’t received a disciplinary write-up in thirteen years, and he was a regular participant in substance abuse and rehabilitative programming. Meanwhile, his most recent psychological evaluation found him to be at low risk of future violence. Because Mr. Alarcon had refused to talk with the psychologist, she only reviewed his prison records and past evaluations. She did not analyze his insight in her report, though his previous evaluation noted that he had “good” insight into the crime.

Back before the board for his sixth subsequent hearing in 2009 following a one-year denial the year before, the commissioners asked for Mr. Alarcon to explain his motives, adding, “You must have had a lot of anger inside you.” Mr. Alarcon said that he could not remember the crime because he had been drinking heavily. However, he related his alcohol use to mounting tension with his wife over unpaid bills and asserted that his crime was not motivated by anger. This led the commissioners to probe deeper into his work and marriage:

PRESIDING COMMISSIONER: All right. What kind of work were you doing at the time of this crime?

INMATE: I was a finisher of -- I was working doing the tiles and the baseboards in the floors in the houses, putting up shelves.

PRESIDING COMMISSIONER: Were you working consistently?

INMATE: Yeah, I had work at a pretty good job there.

PRESIDING COMMISSIONER: Okay. Yeah, you make pretty good money doing that if you’re talented, sure. And so it wasn’t a matter that -- I mean, you know, everybody’s got bills.

INMATE: Right.

PRESIDING COMMISSIONER: Everybody’s just right on the edge it seems, especially now [in 2009], but you had a good job. You and your wife struggling?

INMATE: Yes.

PRESIDING COMMISSIONER: Okay. But no idea why you would cross that line to actually stab this woman, even after she says take the stuff.

INMATE: Well, I believe I blacked out at the time because she was a nice person. I don’t even remember doing it to her.

Here, the commissioner implicitly assessed the potential parolee’s account against what he assumed would lead someone like Mr. Alarcon to violence. In his eyes Mr. Alarcon was working a good job for someone like him—a young, married working-class Latino in the state’s Central Valley—and by the same logic, his relationship woes were not out of the ordinary. Therefore, economic anxiety and relationship instability could not explain his crime.

In the end, the commissioners felt they had no option but to deny Mr. Alarcon parole. In the decision, the presiding commissioner explained:

I thought this would be closer call. You came up a little short on your insight and that seems to be where the struggle is. We need to understand and be confident that you understand why you were an alcoholic, why it would escalate to the point where you would do this particular crime. Not just “I was drunk.” A lot of people get drunk and they don’t cross that line. Otherwise, you’re doing a phenomenal program and you shouldn’t be here.

Crucially, the psychologist did not evaluate his insight and the prior evaluation found him to have good insight. This means that the determination that his insight is lacking solely reflects the opinion of the commissioners. It is also worth noting that Mr. Alarcon did provide an explanation of his crime, related to his marriage and economic struggles. Yet that explanation was rejected out of hand. The commissioner would go on to offer some advice for Mr. Alarcon:

If you would sit down with Title 15 [the BPH regulations], get it out, look at the criteria, look at what we’re looking for. And through Lawrence and Shaputis, the recent cases, we can’t simply deny you based on the crime…. but, you know, when you come in with no insight and just say “I was drunk,” you know, you make it easy for us.

The commissioner told him that based on state regulation, his lack of insight is the last thing keeping him in prison. Yet, as stated earlier, “insight” appears nowhere in state regulation. As indicated by the earlier questioning, Mr. Alarcon’s lack of insight came down to his inability to explain his actions in a way that aligns with commissioners’ implicit assumptions about what would spur someone in his social position to violence. That failure “makes it easy” for the commissioners to deny him parole.

Gaining insight

While insight became part of a vocabulary of denial, commissioners would also occasionally decide that individuals had demonstrated sufficient insight. In the case of Gregory Sheridan, the commissioners explain how they first had to make sense of his social position before they could assess his insight. Mr. Sheridan had a burgeoning acting career when one night he went speeding down a busy LA street after an evening of drinking. He collided head-on with an oncoming car, resulting in the death of the other driver and injuries to several others, in addition to himself.

Since coming to prison Mr. Sheridan had committed himself to multiple rehabilitative programs while receiving no serious disciplinary write-ups. But the board had continued to deny him parole, most recently issuing a one-year denial in 2008. At his fifth subsequent parole hearing in 2009 Mr. Sheridan talked about growing up in a middle-class suburb as the multiracial child of a Black man and a White woman, and how his identity fed his insecurities and his egoism, which led him to think of himself beyond the law. In the decision to grant him parole, the commissioners explained that they found this account compelling. The commissioner elaborated that suspicions about his narcissism and lack of repentance had been problems in prior hearings and particularly in past psychological examinations. Yet, the commissioner went on,

You’ve had lack of insight [sic] that has grown since you’ve been incarcerated. It has increased over the years, which is kind of to be expected, but it doesn’t always. In your case it did, to where we feel now that you do have an adequate amount of insight into why you did what you did, when you did it, concerning your alcohol use, your drug use, perhaps, and driving while under the influence. To say that you’ve had past self-involved or absorbed thinking is something that we discussed at great length in deliberation.

The crux of Mr. Sheridan’s insight lies in his ability to recognize that he was self-absorbed in the past, which led him to act recklessly. However, the commissioner’s determination of insight referenced not only on his words, but also on his self-presentation. The commissioner explained:

Our estimation of you is that you’re well-educated, you’re bright, and you just don’t come cross like a convict, even though you are one. And I think that in many ways, it throws people off. It throws many of the staff off that have looked at you through the years, because we don’t expect to be spoken to by somebody who knows what they’re speaking of and somebody who is fairly successful in their life prior to coming into prison.

In past hearings and evaluations, Mr. Sheridan’s career as an actor led to problems when evaluating his case for parole. Was he sincerely participating in programs, did he sincerely have remorse for the crime, was his commitment so sobriety truthful, or was it all an act? In the mind of the commissioner, this was further complicated by his racial ambiguity: after talking about his education and demeanor, the commissioner stated, “I can say by sitting here looking at you, that you’re a combination of a white mother and a black father. I see both traits. You’re not obvious either one, and you don’t really claim obvious either one. And that’s fine.” The commissioner rearticulated a vision of the prisoner as someone who presents as both not Black and not a criminal. This rearticulation of his social position went hand-in-hand with the commissioner’s assessment of his insight.

Here we see the outlines of insight. Commissioners expect a causal explanation from a potential parolee for their crime. In turn, how well that explanation accords with commissioners’ assumptions about who commits crime and why will justify whether that person gets out of prison. This could stretch across social categories the commissioners saw as relevant to their task. For the working-class Black man who was identified as a gang member, the middle-class White woman who was in an abusive romantic relationship, the unemployed Latino man who was a heroin user, “insight” opened the possibility of commissioners leveraging coarse and essentialist assumptions about why someone like the incarcerated person would join a gang, enter an abusive relationship, or begin using drugs to justify their own determination of insight.

DISCUSSION AND CONCLUSION

In a matter of months following the Lawrence and Shaputis rulings, “insight” became the last hurdle to release for the tens of thousands of people in California prisons serving indeterminate life sentences. The term was once widely used in parole decision making across the United States in the middle of the twentieth century, during the height of rehabilitative approaches to incarceration. But by 2007 at the BPH it survived as a secondary feature of supplemental psychological assessments that commissioners reviewed and referenced to support their decisions. When the Lawrence and Shaputis rulings declared commissioners’ primary justification for denying parole to an individual—the heinousness of the crime—out of compliance with legal requirements for their decision making, commissioners uprooted “insight” from the psychological assessments and moved it to the center of their decisions. With this, commissioners increasingly referenced not expert judgment from psychologists but rather their own judgment about a potential parolee’s insight. The refurbished term came to constitute a new vocabulary of denial grounded in old rehabilitative language, while it created possibilities for commissioners’ decisions to be anchored in assumptions based on an incarcerated person’s race, class, and gender.

The emergence of “insight” at the center of decisions to incarcerate individuals appears surprising in the face of predominant understandings of contemporary penal practice that emphasize shifts toward risk and retribution. These accounts suggest that “insight” would have no place in a landscape of punishment dominated by risk and retribution. So how did it come back to the foreground of a major penal institution in the twenty-first century?

Toward this end, the article supports and extends recent research arguing that the turn to retribution and risk is only partial. Penal organizations accumulate a broad array of disparate, and sometimes seemingly contradictory, practices over time (Lynch 2000; Goodman 2012; Werth 2013; Goodman, Page, and Phelps 2017). These practices, lingering on the edges of administrative processes, can keep old and outmoded ideas alive in the routine functioning of penal organizations. To understand how this shapes bureaucratic change, I argue that we should turn to the dynamics of compliance (Edelman 1992; Mahoney and Thelen 2010; Lara-Millán 2021).

In this case, when the dominant idea of punishment guiding parole board decision making was suddenly ruled out of compliance by the court, “insight” was readily accessible for commissioners. Focusing on the dynamics of compliance further suggests why “insight” specifically. First, in line with what scholars of legal endogeneity would expect (Edelman 1992; Edelman, Uggen, and Erlanger 1999; Dobbin and Kelly 2007; Edelman et al. 2011), the term leaned into expertise and tradition to signal compliance with the court and avoid further litigation. While not scientifically validated as a means of determining parole release for life-sentenced prisoners, the term had a historical record of social validation in penal contexts and expert validation through the psychological assessments. At the same time, “insight” also met the demands of a second source of compliance: the governor. “Insight” meshed snugly with the governor’s professed interest in rehabilitation, while preserving a great amount of discretion.

More generally, the article suggests refocusing analyses of bureaucratic penal change on the dynamics of compliance. Compliance defines a set of relevant actors for analyzing administrative change: those who can directly intervene in administrative practice in the name of compliance. Compliance in this sense can come from a wide range of actors, including courts, elected officials, funding sources, and oversight committees, each in different ways and with differing powers to define compliance. Further, it focuses attention on concrete, routine administrative practices that are in or out of compliance, rather than the professed ideological commitments of administrators, which may or may not align with their practices.

A focus on compliance translates directly into a set of implications for understanding longevity over time and heterogeneity across place in penal practice. This would imply that the penal organizations most thoroughly remade by the turn toward risk and retribution are those where (a) the mechanisms for enforcing compliance were strongest, and (b) the actors defining compliance were most affected by the broader social, economic, and political conditions that fostered the turn to risk and retribution. Meanwhile, parts of the criminal-legal system where the possibility of intervention in the name of compliance is least would be most insulated from changes driven by broader political, social, and economic trends.

While the article is not designed to test these implications, they should be investigated in future research. Given the research design, the article can also only show the possibility of commissioners’ common sense being grounded in racial, class-based, and gendered assumptions about criminality. This point, too, deserves further investigation to elaborate the extent and contours of this penal common sense.

Meanwhile, the article shows how penal institutions are remarkably robust in their ability to achieve compliance through refurbishing old practices and ideas. Scholars of legal endogeneity argue that private organizations satisfy new definitions of compliance through the development of new practices, guided by the “rational myths” of outside experts (Edelman, Uggen, and Erlanger 1999; Dobbin and Kelly 2007). In contrast, penal bureaucracies have accrued a wide array of disparate and seemingly contradictory practices over decades that carry the mythical weight of tradition and expertise and thus are readily available to meet changing definitions of compliance. Under these conditions, compliance can be achieved through refurbishing existing practices rather than the development of new practices. This framework further predicts that refurbishing old practices is most likely where an outside entity redefines compliance in a way that demands a rapid change to organizational routines.

Finally, the article has practical implications. Much has changed since the 2008 court rulings. In 2007, 1.9 percent of scheduled hearings resulted in a parole grant. In 2021, 16.3 percent did. ¹¹ The number of people granted parole has slowly increased amid changes in governors, new parole board leadership and efforts to diversify the professional background of commissioners, new procedures to formalize the hearing process, a US Supreme Court decision requiring the state to significantly decrease its prison population, a widening number of people eligible to go before the parole board, and a broader (if uneven) shift in popular understandings of the role of incarceration in society.

Current observers and practitioners suggest that what has not changed is the centrality of “insight” to the reasoning of parole decisions (Wattley 2013; Shammas 2019; Aviram 2020; Greene and Dalke 2021; UnCommon Law 2022). Incarcerated people now must go through an immensely challenging process of finding a credible way to convince commissioners of their insight in hearings, often with insufficient support and with little guarantee for their efforts (Aviram 2020; Petek 2023).

The ultimate consequence of the court rulings was not to constrain commissioner discretion as much as to reroute it through a new language of justification. If taken literally, the understanding of “insight” put forward in hearings would suggest that mass incarceration was driven by the growth of people who lack understanding of their inner motivations. This is at odds with the scholarly consensus that US prisons filled with primarily poor Black and Latino men over the last fifty years because of increasingly punitive and racially targeted politics and policies (Garland 2001; Wacquant 2009; Alexander 2010; National Research Council 2014; Hinton 2016). These political decisions—including decisions about neighborhood and school investment, the economic safety net, and health care provision—disappear in accounts that seek individualized answers to the question: why did you do this?

This suggests the limits of attempts to achieve reform through the proliferation of external compliance mechanisms for individual criminal-legal organizations, from police departments to criminal courts to prisons. Instead, more durable change may be the product of establishing and scaling up new organizations loosened from the inertia of traditional routines and accumulated practices. In other words, the most important “breaks” or “ruptures” in the practices of punishment might not be epistemic or ideological, but organizational. If this is correct, the growth of new organizational configurations today, such as those emerging in the realms of restorative justice and violence prevention, will leave a longer legacy than the refurbishment of existing criminal-legal bureaucracies. Yet as the California parole board suggests, such organizations grow to contain their own befuddled set of contradictory practices amid ambivalent and ever-changing dynamics of compliance.