1.1 Neurolaw: A Brief Introduction

The 1990s were designated the “decade of the brain” by President George H.W. Bush and the US Congress. This period was marked by an extraordinary increase in the visibility of neuroscience and a renewed interest in the localization of neural phenomena to distinct anatomical regions of the brain. The rapid growth of functional magnetic resonance imaging (fMRI) research, for instance, led to new insights into neuroanatomical structure and function, which, in turn, led to a greater understanding of human behavior and cognition. In response to these developments, questions emerged regarding how these findings can be applied to criminology and legal processes. As Martin Roth describes, “The last two decades have seen an explosion of interest in what impact, if any, our growing knowledge of the brain will or should have for legal theory and practice” (Reference Roth2018: 1). The field of study that has developed around this interest is now commonly referred to as “neurolaw.”

Neurolaw is an interdisciplinary field that involves collaboration between neuroscientists, legal scholars, psychologists, and ethicists. Its goal is to bridge the gap between neuroscience and the law, providing a scientific foundation for informed legal decision-making. By “neuroscience,” I include not only those areas of research that study the brain directly (its structures, functions, developments, and abnormalities), but also those that study behavioral genetics and cognition more generally. I, therefore, interpret the “neuro” part of neurolaw rather broadly.

The term “neurolaw” was first coined by Reference Taylor, Harp and ElliotTaylor, Harp, and Elliot (1991) in a Neuropsychology journal article analyzing the role of psychologists and lawyers in the criminal justice system. After its publication, scholars began networking, organizing conferences, and publishing books and articles about the intersection of neuroscience and law. The developing field of neurolaw was then given a boost by the initiation of the Law and Neuroscience Project by The MacArthur Foundation. Phase 1 of this project was launched in 2007 with a $10 million grant. The initiative sustained forty projects addressing a multitude of issues, including experimental and theoretical data that explored how neuroscience may eventually shape the law. Since then, a number of universities have developed neurolaw centers dedicated to exploring the social, legal, and ethical implications of neuroscience.

Neurolaw is closely aligned with, and developed in parallel with, neuroethics (see Reference RoskiesRoskies 2002, Reference Roskies2021; Reference LevyLevy 2007, Reference Levy2012). Neuroethics is a sub-discipline of philosophy with two broad focuses. The first, which has come to be called the ethics of neuroscience, concerns the assessment of ethical issues arising from neuroscience, its practice, and its applications; the second, which has come to be called the neuroscience of ethics, concerns the ways in which the sciences of the mind can illuminate longstanding issues within philosophy (Reference LevyLevy 2012: 143). The ethics of neuroscience deals, for instance, with concerns about the application of fMRI-based lie detection systems in forensic settings and ethical considerations raised in the course of designing and executing neuroscientific studies. The neuroscience of ethics, on the other hand, deals with the degree to which we might be required to revise folk or philosophical conceptions of the self in the light of work in cognitive neuroscience. As Adina Roskies explains, “Traditional ethical theory has centered on philosophical notions such as free will, self-control, personal identity, and intention. These notions [however] can [now] be investigated from the perspective of brain function” (Reference Roskies2002: 22). This, Roskies predicts, will have “profound implications for the way ethics, writ large, is approached in the 21^st century” (Reference Roskies2002: 22).

While there is some controversy over whether neurolaw should be seen as a branch of neuroethics, the two have sufficiently distinct areas of inquiry to justify independent treatment. Gerben Meynen writes, “[Neurolaw] could be considered merely a branch of neuroethics, but that would not do justice to its central objective, which is examining the impact of neuroscience on the law” (Reference Meynen2014: 819). What differentiates neurolaw is that, in general, it “deals in much more legal detail with the issues raised by neuroscience than is done in neuroethics.” Whereas neuroethics might investigate ethical questions about the use of brain-based lie detection, neurolaw considers questions like: given the specific legal tests for admissibility of evidence in the United States, under what circumstances could brain-based lie detection be used in legal cases? Given its focus, then, on the meaning and implications of neuroscience for the law and legal practice, I maintain that whether or not neurolaw remains, in part, a branch of neuroethics, it is unique enough to deserve its own treatment.

1.2 Assessment, Intervention, and Revision

While there are different ways to go about classifying the different areas of neurolaw, I find it most helpful to divide them into three basic domains of investigation: assessment, intervention, and revision (Reference Morse and RoskiesMorse and Roskies 2013; Reference MeynenMeynen 2014). The first domain, assessment, concerns evaluations of the state of mind/brain of individuals, including defendants, prisoners, and jurors. For instance, in the future, people’s brains may be analyzed using neuroscientific techniques to answer questions about lying, insanity, risk of recidivism, or bias (Reference Greely, Morse and RoskiesGreely 2013; Reference MeynenMeynen 2014). With respect to such assessments, “the reliability and validity of neuroscience technique, not just in research setting but specifically in criminal law cases, is a central topic of debate” (Reference MeynenMeynen 2014: 820). It is in this domain that neurolaw investigates legal issues related to risk assessment for future violence, assessment of criminal responsibility or legal insanity, brain-based mind reading like evaluating biases in prospective jurors, and measuring deviant sexual appetites in defendants.

Consider, for instance, the mental state requirement in criminal cases and how it gave birth to forensic psychology and other methods of assessment. In most cases, for a person to be convicted of a crime, the prosecutor must prove that the defendant committed an unlawful act (the physical element) and that they had the intention to commit the unlawful act (the mental element). The former is referred to as actus reus, the Latin phrase for “guilty act.” Actus reus is the wrongful deed that comprises the physical component of a crime. In most criminal cases, actus reus must be coupled with mens rea – a “guilty mind” – for one to be held criminally liable. Mens rea represents the defendant’s state of mind at the time of the crime, and the prosecution must prove that the defendant had the mental element of “guilty mind” while committing a crime to secure a conviction. There are, however, a number of excuse defenses that can be used to show that a defendant’s mental state demonstrates that they should not be held legally responsible for a criminal act. Such excuses include insanity, diminished capacity, duress, mistake, and infancy.

Building on this, Nicole Vincent writes:

It is in this context that advances in the behavioral, cognitive, and neurosciences might again affect legal responsibility practices by “providing further insights into the nature of human agency and by offering revamped diagnostic criteria and more powerful diagnostic and intervention tools with which to assess and to alter minds” (Reference Vincent and VincentVincent 2013: 2–3). Neurolaw therefore seeks to investigate the potential benefits of using powerful new diagnostic brain imaging techniques to help assess the mental states, disorders, and mental capacities of defendants.

The second domain of neurolaw concerns all types of brain-based and law-related interventions. As Meynen explains, “In this domain, brains are not so much analyzed or interpreted, but changed” (Reference Meynen2014: 821). In particular, he highlights three types of interventions: treatment, enhancement, and manipulation.

Each type of intervention raises its own unique issues. For instance, as our understanding of the brain increases, there has been growing support for neuroscientifically informed rehabilitation programs, which hold out the promise of reducing crime, increasing safety, and helping treat those offenders who suffer from mental health problems, addiction, pedophilic desires, and the like. These different types of interventions are made possible by new neuroscientific techniques and technologies – for example, transcranial magnetic stimulation, neurofeedback therapy, deep brain stimulation, and improved psychopharmaceuticals. But the opportunities they create – to treat, to enhance, and generally to modify people’s minds – raise a number of vexing and important issues. For example, do such interventions violate the conditions for legally effective informed consent, alter an agent’s personal identity in philosophically problematic ways, or otherwise violate an agent’s rights? These are just some of the questions neurolaw investigates within the domain of intervention.

The third and final domain, revision, concerns the extent to which neuroscientific findings require us to revise, or even eliminate, parts of the law. Consider, for instance, the criminal prosecution of children. Currently, over half of the states in the United States still have no minimum age for prosecuting children, leaving eight-, nine-, and ten-year-old children vulnerable to extreme punishment, trauma, and abuse within adult jails and prisons. This has led to prosecutors across the country prosecuting young children in adult court. While the United States may be an outlier in the practice of prosecuting young children, it is by no means alone. Australian law also currently allows children as young as ten to be charged with a criminal offence, falling below the most common minimum age of criminal responsibility worldwide of 14. Children, however, are especially immature and impulsive and have not yet developed mature judgment or the ability to accurately assess risks and consequences. A neuroscientific understanding of adolescent brains could help us decide whether our current legal practices need to be revised (see Reference Hirstein, Sifferd and FaganHirstein, Sifferd, and Fagan 2018).

Another area where revision is possible, though far more controversial, has to do with the law’s assumptions about free will and moral responsibility and whether they can be reconciled with a neuroscientific understanding of human agency. Some theorists, for instance, maintain that neuroscience shows that free will is an illusion. Therefore, nobody is truly responsible for their actions, hence the criminal law has to be revised (Reference Green and CohenGreene and Cohen 2004; Reference AlcesAlces 2018). Others maintain that neuroscience is completely irrelevant to criminal law and, hence, no major revisions are necessary (Reference MorseMorse 2005, Reference Morse and Vincent2013, Reference Morse2015, Reference Morse, Caruso and Flanagan2018). This aspect of neurolaw can be framed as a concrete instance of the challenge of fusing what Wilfrid Sellars called the manifest and scientific images of human beings. As Roth explains:

While some neurolaw practitioners think a fusion of the manifest and scientific images is possible (Reference Hirstein, Sifferd and FaganHirstein, Sifferd, and Fagan 2018; Reference RothRoth 2018), others disagree (Reference Green and CohenGreene and Cohen 2004; Reference AlcesAlces 2018). For instance, Joshua Greene and Jonathan Cohen have famously argued that neuroscience will motivate us to re-examine criminal legal doctrines and penal policies that presuppose wrongdoers can be morally responsible for their actions in the desert-based sense:

Whether or not they are correct about this is something I will explore further in Section 4, but note how the last sentence captures both the descriptive and normative aspects of what neurolaw aims to investigate – that is, Greene and Cohen make both a prediction about how neuroscience will impact the legal system as well as a normative endorsement for revision. Given that how things are (or will be in the future) can be distinct from how they ought to be, they can be right about one, both, or neither of these claims. Hence, neurolaw investigates both types of questions.

1.3 Metaphysical, Moral/Social, and Evidential Issues

In addition to the three domains of neurolaw just outlined, we can also say that within each, the kinds of issues that arise typically fall into one or more of the following categories: metaphysical, moral/social, and evidential (Reference RothRoth 2018: 8). Foremost among the metaphysical issues relevant to neurolaw are the nature of human beings, the mind-body problem, and the question of free will (Reference RothRoth 2018: 8). Consider, for instance, the case of Charles Whitman. By all accounts, Whitman was a smart, talented, and popular young man – a paradigm of the “all-American boy.” He was an Eagle Scout, ex-marine, and promising engineering student. Yet on August 1, 1996, Whitman did the unthinkable. He killed his wife and mother and then went on a shooting spree at the University of Texas at Austin. He fatally shot three people inside UT Austin’s Main Building and then ascended to twenty-eighth-floor observation deck on the building’s clock tower. There, he fired at random people for ninety-six minutes. In the end, he killed sixteen people and wounded thirty-two others before being shot to death by police. Months early, however, Whitman began complaining of severe, persistent, and painful headaches. He also confessed in his journal of “being the victim of many unusual and irrational thoughts.” In his suicide note, Whitman requested an autopsy be done to examine his brain since he was convinced it would show some “visible physical disorder.” The autopsy confirmed Whitman’s suspicions. It revealed a large tumor in Whitman’s brain which had impacted the amygdala, a structure of the brain implicated in regulating emotions, including fear and aggression.

While Whitman’s case is disturbing for a number of reasons, it is by no means unique. In fact, neuroscientists are beginning to shed new light on how brain lesions can lead to criminal behavior. A recent study conducted by Reference Darby, Horn, Cushman and FoxDarby et al. (2017) systematically reviewed seventeen known cases where criminal behavior was preceded by the onset of a brain lesion. They found that while the lesions were distributed throughout different brain regions, all the lesions were part of the same functional network located on a single circuit that normally allows neurons throughout the brain to cooperate with each other on specific cognitive tasks. The network identified by the researchers is closely related to networks previously linked with moral decision-making. In particular, the network is most closely associated with two specific components of moral psychology: theory of mind and value-based decision-making (Reference Darby, Horn, Cushman and FoxDarby et al. 2017). Theory of mind refers to the capacity to understand other people’s points of view, beliefs, and emotions. This helps you appreciate, for instance, how your actions would make another person feel. Value-based decision-making refers to the ability to judge the value of specific actions or their outcomes. This helps you see not only what the outcome of your actions will be, but whether those actions and outcomes are good or bad.

Cases like Whitman’s highlight just how important metaphysical questions are to the criminal law. Did Whitman’s brain tumor make him go on a killing spree? If so, did he have the kind of control in action (i.e., free will) required for desert-based moral responsibility when he committed his crimes? More generally, as our predictive grasp of complex behavior improves, how will the bolstered sense that the mind is identical with, supervenient on, or emergent from the brain affect or undermine our notions of free will or moral responsibility? What if everything we do is the result of our neurobiology? And what if that neurobiology were the result of the same deterministic laws of nature that dictate the behavior of all other physical things in the physical universe? How we answer these questions will have a major impact on the kinds of issues investigated by neurolaw, especially with regard to the domain of revision.

These metaphysical questions also have direct relevance to neurolaw in connection with the second category of moral/social issues. The problem of free will, for instance, is intimately connected with our moral and legal practices and institutions. In fact, most contemporary philosophers simply define free will as the control in action required for a particular but pervasive sense of moral responsibility. This sense of moral responsibility is typically picked out by the notion of basic desert (see Reference PereboomPereboom 2001, Reference Pereboom2014; Reference Caruso and MorrisCaruso and Morris 2017; Reference CarusoCaruso 2021a; Reference Caruso and PereboomCaruso and Pereboom 2022). As Derk Pereboom defines it:

Understood this way, free will is a kind of power or ability that an agent must possess in order to justify certain kinds of basically deserved attitudes, judgments, and treatments – such as resentment, indignation, moral anger, and retributive punishment. These reactions would be justified on purely backward-looking grounds; that’s what makes them basic and would not appeal to consequentialist or forward-looking considerations such as future protection, future reconciliation, or future moral formation.

By defining free will as the control in action required for basic desert moral responsibility, it’s easy to see the practical importance of the traditional debate to our moral and social lives. If it turns out, for instance, that agents lack the kind of free will required for basic desert moral responsibility, then no one is deserving of praise and blame in the basic sense (Reference PereboomPereboom 2001, Reference Pereboom2014, Reference Pereboom2021; Reference CarusoCaruso 2021a). Adopting this skeptical position would also undermine the retributive justification of legal punishment, since it does away with the idea of basic desert altogether – that is, if agents do not deserve blame just because they have knowingly done wrong, neither do they deserve punishment just because they have knowingly done wrong (Reference PereboomPereboom 2014: 157; Reference CarusoCaruso 2021a). For this reason, neurolaw concerns itself not only with the moral/social/legal implications of the traditional free will debate, but also with the aims of our criminal justice system, including the practice of holding people criminally responsible and the underlying rationale for adopting the legal definitions and distinctions that inform judgments of legal responsibility.

The final category concerns evidential issues, such as the methods or procedures by which we come to know about the biological and psychological states of people (Reference RothRoth 2018: 9). These issues most closely align with the domain of assessment, especially when brain scans are used to assess defendants and then introduced during court proceedings as, say, exculpatory evidence of brain damage or insanity. Here, we must ask: What do these brain images purport to show? How were they produced? And how reliable is the information provided by them?

Neurolaw, therefore, investigates whether brain-based methods of assessment should be introduced in court proceedings and, if so, what evidential weight they should be given.

Bringing everything together, we can now say that the domains of assessment, intervention, and revision, together with the metaphysical, moral/social, and evidential issues addressed by neurolaw, yield the classification of neurolaw topics shown in Table 1:Footnote ²

2.1 Assessing Competency, Guilt, and Sentencing

From the introduction of the electroencephalograph (EEG) in the 1930s to the first magnetic resonance imaging (MRI) scans performed on humans in the 1970s, the twentieth century saw great advances in neuroscience and neuroimaging specifically. “These tools not only gave scientists an inside view into the structure and function of the human brain, but they also allowed experts to better conceptualize the connection between the human brain and human behavior” (Reference Aono, Yaffe and KoberAono, Yaffe, and Kober 2019: 2). This connection has become particularly relevant in the courtroom since the more we understand about neuroscience, the more we see that even subtle abnormalities in the brain can affect moral reasoning and human behavior, which in turn can affect competency and blameworthiness. As a result, the use of neuroscientific evidence in criminal proceedings has increased significantly over the last two decades.

Reference ShenFrancis Shen (2016) traces one of the earliest introductions of neuroscience into courtrooms to the 1940s when EEG was first used in a case involving a defendant with epilepsy. As new neuroimaging technologies emerged, they too made their way into the courtroom. In 1981, John Hinckley’s attempted assassination of president Ronald Reagan led to one of the highest-profile cases that utilized neuroscience in a criminal trial. Hinckley’s defense team introduced a computerized axial tomography (CAT) scan to help bolster their arguments that he suffered from schizophrenia due to a “shrunken” or atrophied brain and should therefore be found not guilty by reason of insanity. Although the prosecution opposed the introduction of Hinckley’s CAT scans as evidence, arguing that it did not prove he had schizophrenia, it was ultimately allowed since the sort of brain atrophy he had was more common among schizophrenics than among the general population. Ultimately, the scans helped convince the jury, and they found Hinckley not responsible by reason of insanity (Reference Aono, Yaffe and KoberAono, Yaffe, and Kober 2019: 2).

A decade later, a new form of neuroimaging made an appearance in People v. Weinstein (1992). Herbert Weinstein was charged with second-degree murder for strangling his wife and throwing her from the 12^th floor of their Manhattan apartment, a charge he readily admitted to.

The arachnoid cyst was situated within the left sylvian fissure and compressed the left frontal, temporal, and insular regions of Weinstein’s brain. Weinstein’s attorney offered the PET scan in support of a claim of not guilty by reason of insanity. The prosecutor, on the other hand, moved to preclude any testimony or other evidence concerning the PET scan – arguing that PET scans were not accurate or reliable depictions of cerebral metabolism. While there was no doubt that the presence of the cyst altered Weinstein’s brain structure and function, the question before the court was one of causation. As Susan Rushing writes: “Was there sufficient evidence to allow psychiatric and neurological experts to testify that Mr. Weinstein’s brain abnormality was related to his violent criminal behavior?” (Reference Rushing2014: 63). Unfortunately, we have no way of knowing how Weinstein’s insanity claim would have fared before a jury because on the eve of trail he agreed to plead guilty to the lesser charge of manslaughter and was sentenced to seven to twenty-one years in prison.

In more recent decades, since around the turn of the century, the use of neuroscientific evidence in criminal trials has continued to increase (Reference FarahanyFarahany 2016; Reference MeixnerMeixner 2016). In a study of US cases between 2005 and 2012, Reference FarahanyNita Farahany (2016) found that “the use of neurobiological evidence by criminal defendants is increasing year over year” (Reference Farahany2016: 491). In particular, she found that 1585 judicial opinions from criminal cases discussed the use of neurobiological evidence by criminal defendants to bolster their criminal defense. In 2012 alone, there were 250 judicial opinions written in which the criminal defendant argued (successfully or otherwise) that their “brains made them do it” – more than double the number in 2007. In fact, Farahany found that 5 percent of all murder trials and 25 percent of all death penalty trials in 2012 featured criminal defendants making a bid for lower responsibility or lighter punishment using neurobiological data (Reference Farahany2016: 486; see also Reference DennoDenno 2015).

Farahany’s study also revealed additional surprising results and trends. While many scholars have discussed the implications of using neurobiological evidence for mitigating criminal punishment, Farahany found that the second most common use of neurobiological evidence in criminal cases is to challenge competency. She also found that the quality, and not just the quantity, of opinions discussing neurobiological evidence has evolved. Opinions earlier in the study often discussed neurobiological evidence as part of a laundry list of other types of scientific evidence introduced. In later opinions, however, “judges spilled substantial ink discussing the neurobiological evidence often in significant detail and with citations to scientific literature and the experts who testified in the case” (Reference Farahany2016: 492). According to Farahany, this suggests “a shift in both the frequency and the nature of how such evidence is being evaluated by judges and juries in criminal cases” (Reference Farahany2016: 492).

Despite this shift, important questions and criticisms remain. Critics, for instance, point out that the use of neurobiological evidence has thus far been rather “haphazard, ad hoc, and often illconceived” (Reference FarahanyFarahany 2016: 488–9). This is because, despite the caution and nuance neuroscientists typically advise, prosecutors and defense attorneys are often quick to introduce neurobiological evidence and sometimes even overstate it whenever they think it can help their case. Defense attorneys introduce neuroscientific evidence in attempts to exculpate criminal defendants, to bolster preexisting legal defenses, and to mitigate defendants’ culpability and punishment. Prosecutors, on the other hand, have “seized upon the double-edge potential of a claimed neurobiological evidence to denigrate defendants’ characters and to demonstrate defendants’ likely future dangerousness” (Reference FarahanyFarahany 2016: 489). Given the gravity of such decisions, including assessments bearing on deservingness for capital punishment, it is important that we separate the wheat from the chaff and investigate the proper role of neuroscience in criminal proceedings.

While the interconnection between neuroscience and law is promising and rapidly developing, there are several critical methodological cautions we need to keep in mind. First, studies of how an average brain works do not always provide reliable information on how a specific individual’s brain works. Functional imaging research typically relies on results based on population-level inferences, yet group average results may belie the unique patterns of activity present in the individual. Second, “there are many types of brain imaging techniques, with many accompanying ways to be interpreted and presented” (Reference DuDu 2020: 511). For example, PET scans use radioactive tracers to detect blood concentrations in different brain regions that are associated with brain functions. EEGs record electrical activity in the brain, which is then analyzed with the aid of a computer system to allow for inferences about brain function. And functional MRI detects changes in hemodynamic properties of the brain that typically occur when engaging in particular mental tasks. As Yu Du writes: “It is vital to bear in mind that brain images are the result of a process within a process. In other words, many decisions and steps are involved in determining exactly when and what data should be collected and how the data should be analyzed and presented. The room for technical and statistical mistakes and misinterpretations can be significant” (Reference Du2020: 512).

Another concern is that the meaning of brain images is not straightforward or self-evident. As such, inferring the psychological, moral, and/or legal significance of such images is not easy. We cannot detect mens rea or moral responsibility on a brain scan. “Even well-designed, well-executed, properly analyzed, and properly presented brain images must be interpreted in the correct context” (Reference DuDu 2020: 513). Furthermore, identifying structural abnormalities in the brain and correlating them with violent behaviors is “complicated and subject to enormous variation” (Reference DuDu 2020: 513). This is because “many brain regions are involved in a wide variety of functions” (Reference DuDu 2020: 513). Additionally, the same task can also be accomplished in the presence of widely varying brain activation patterns. To make matters worse, even when a strong correlation can be established, it’s generally not enough to prove causation. Nor is it always clear what it can tell us about competency, intent, and the state of mind of the individual at the time of the crime. It’s possible, for instance, for neuroimaging to confirm that structural changes have occurred in the brain (due, say, to a brain injury or tumor) and perhaps even that this resulted in behavioral change. What it cannot do – and, hence, what is left for neurolaw practitioners to debate – is determine whether such functional and behavioral changes should reduce culpability or mitigate punishment. In some cases, it can be rather difficult to determine how much weight we should give to the neurobiological evidence and what, if anything, it can tell us about an individual’s competency, guilt, and/or punishment.

I believe it’s important, then, that when discussing the use of neuroscientific evidence, we keep all these methodological limitations in mind. Acknowledging these limitations, however, does not mean we should prohibit the use of neuroscience in court. To do so would be to throw the baby out with the bathwater. I contend that when used responsibly, neuroscientific evidence has the potential to improve the accuracy and decrease errors in the criminal justice system. I therefore agree with Farahany when she writes, “decrying the use of neurobiological evidence in criminal law seems both futile and counterproductive; neuroscience is already entrenched in the US legal system. And used appropriately, it holds promise of improving decision-making in law. An outright ban is neither warranted nor productive” (Reference Farahany2016: 488). Rather than resisting this shift, then, neuroscientists should play a part in safeguarding these developments by helping educate the public and legal professionals about the responsible use of neuroscience in the courtroom.

One area where brain scans and neuroscientific evidence may be of value is in cases where a tumor or traumatic brain injury caused damage that resulted in a significant personality change. Here, “the role of neuroimaging is important to the case as it can provide evidence for a reason behind the alteration of their personality” (Reference Straiton and LakeStraiton and Lake 2021: 70). Alteration in personality due to a structural change in the brain can include increased impulsiveness, depression, aggression, inappropriate sexual behavior, lack of thought control, and violence. Consider, for instance, the widely discussed case of a forty-year-old schoolteacher who experienced new-onset pedophilia due to a right orbitofrontal brain tumor – not once, but twice.

According to Reference Burns and SwerdlowJeffrey Burns and Russell Swerdlow (2003), the two neurologists who first treated the man, his behavior changed significantly in the year 2000. While the individual had a preexisting interest in pornography, he increasingly began to frequent Internet pornography sites that “emphasized children and adolescents and was specifically targeted to purveyors of child pornography” (Reference Burns and Swerdlow2003: 437). The man also began soliciting prostitution at “massage parlors,” which he had not previously done.

As a result, the judge sentenced him to prison. However, the night before entering prison, the man complained of severe headaches and was taken to the emergency room where, following an fMRI scan, it was revealed that he had a right orbitofrontal tumor. The tumor was removed, his behavior returned to normal, and he was deemed safe to go home to his wife and stepdaughter. Several months later, the sexually charged behavior returned, and it was discovered that the tumor had regrown. It was again removed, and once again his behavior returned to normal.

Since the man’s symptoms were resolved each time the orbitofrontal hemangiopericytoma was removed, it seems reasonable to infer a causal connection between the man’s behavioral changes and orbitofrontal disruption in his brain. In fact, the two neurologists on the case themselves see this as “further establishing causality” (Reference Burns and SwerdlowBurns and Swerdlow 2003: 439). According to them, the orbitofrontal disruption likely exacerbated a preexisting interest in pornography, manifesting as sexual deviancy and pedophilia. To their knowledge, “this is the first description of pedophilia as a specific manifestation of orbitofrontal syndrome” (Reference Burns and Swerdlow2003: 439). They also go on to conclude that: “Our patient could not refrain from acting on his pedophilia despite the awareness that his behavior was inappropriate” (Reference Burns and Swerdlow2003: 440). This conclusion, they claim, is supported both by the details of the case and the fact that “Orbitofrontal lesion research suggests that sociopathic behavior results from a loss of impulse control rather than a loss of moral knowledge” (Reference Burns and Swerdlow2003: 439). In their professional opinion, the tumor interrupted connections between the orbitofrontal lobe and the amygdala – the region of the brain responsible for emotion and decision-making – therefore resulting in diminished impulse control.

This assessment seems to fit with everything we currently know about the orbitofrontal cortex. In fact, the orbitofrontal cortex is one region that has been consistently associated with antisocial or violent behavior, along with the anterior cingulate cortex and the dorsolateral prefrontal cortex – both also located in the frontal lobe (Reference Yang and RaineYang and Raine 2009; Reference Sajous-Turner, Anderson and WiddowsSajous-Turner et al. 2020; Reference Straiton and LakeStraiton and Lake 2021). Neuroimaging studies have also found that patients who suffered injury to these regions of the prefrontal cortex show reduced decision-making capabilities and psychopathic-like behavior (Reference Blair and CipolottiBlair and Cipolotti 2000; Reference Straiton and LakeStraiton and Lake 2021). Such research not only supports the neurologists’ assessment in the new-onset pedophilia case, but it also highlights how neuroscience can be used in court to show how brain lesions and injuries – particularly ones to the frontal lobe – can radically reduce the ability to make rational decisions and check impulses.

Another area where neuroscientific evidence can be helpful is in determining punishment, especially in death penalty cases. On a case-by-case basis, there is unfortunately no hard and fast principle for determining when and to what extent neuroscientific evidence should lead a jury to a non-guilty verdict. As a result, in many cases, judges and juries may not find neuroscientific evidence convincing enough to completely absolve a defendant. However, if the guilty verdict is finalized and the decision is between a life sentence and the death penalty, neuroscientific evidence (along with other evidence about the impact of early life experiences, genetics, and epigenetics) could be relevant (Reference DuDu 2020: 501).

During the penalty phase of capital trials, defendants may introduce mitigating evidence that argues for punishment less than death. Brain scans, for instance, can be used to identify and support defendants’ claims of diminished culpability due to circumstances beyond their control in capital cases (Reference Focquaert, Glenn, Raine and NadelhofferFocquart, Glenn, and Raine 2013). Defendants may also introduce neuroscientific evidence of brain damage or dysfunctionality in an attempt to mitigate punishment. In fact, empirical research has shown that scientific evidence about a hypothetical defendant’s brain function is likely to limit a sentence’s severity. A study conducted in Science by Reference Aspinwall, Brown and TaberyAspinwall, Brown, and Tabery (2012) found that, on average, judges subtracted a year from an imaginary convict’s sentence after being told he was genetically predisposed to violence. And Reference Green and CahillEdith Greene and Brain Cahill (2012) conducted a study assessing the impact of neuroscientific evidence on mock jurors’ sentencing recommendations and impressions of a capital defendant. Using actual case facts, they manipulated diagnostic evidence presented by the defense (psychosis diagnosis, neuropsychological test results, and neuroimages) and found that defendants were less likely to be sentenced to death when jurors had this evidence than when they did not. They also found that neuropsychological and neuroimaging evidence had mitigating effects on impressions of the defendant.

While some theorists claim that neuroscientific evidence is a “double-edged sword” – one that will either get defendants completely off the hook or unfairly brand them as posing a future danger to society – a recent empirical analysis by Reference DennoDeborah Denno (2015) reveals this to be a myth. Her study evaluated 800 criminal cases involving neuroscientific evidence over the past two decades. It was the first to empirically and systematically investigate how courts assess the mitigating and aggravating strength of such evidence. Her analysis found, first, that neuroscientific evidence is “usually offered to mitigate punishments in the way that traditional criminal law has always allowed, especially in the penalty phase of death penalty trials” (Reference Denno2015: 493). Contrary to the popular image of neuroscience as a double-edged sword, her study found that “neuroscience evidence is typically introduced for a well-established legal purpose – to provide fact finders with more complete, reliable, and precise information when determining a defendant’s fate” (Reference Denno2015: 493). Second, her study also found that “courts accept neuroscience evidence for this purpose, and in fact expect attorneys to raise this evidence when possible on behalf of their clients” (Reference Denno2015: 494). In fact, this expectation is so entrenched that “courts are willing to grant defendants their ‘ineffective assistance of counsel’ claims when attorneys fail to pursue this mitigating evidence” (Reference Denno2015: 495). Denno’s study also found that “the potential future danger posed by defendants is rarely a facet of cases involving neuroscience evidence – again contradicting the myth of the double-edged sword” (Reference Denno2015: 495). Lastly, her study found that neuroscience evidence usually mitigates punishment, particularly for death penalty sentences. It seems, then, that neuroscientific evidence may be more influential at the sentencing phase than the liability (guilt or innocence) phase and is more likely to mitigate punishment than exacerbate it.

Instead, then, of entirely preventing the use of neuroscience in the courtroom to assess competency, guilt, and sentencing, I contend that its application should be careful, specialized, and context-specific. We must begin by acknowledging the methodological limitations outlined previously, including concerns about statistical significance, effect size, and the difficulty of making legal and psychological inferences. We must also acknowledge that the application of neuroscience in the courtroom is highly context-specific. That said, neuroscientific evidence is valuable “not because it creates something new but instead because it explains the normative behaviors and folk-psychological concepts within the legally relevant domain” (Reference DuDu 2020: 515). When properly used, neuroscience can help us make more scientifically informed decisions within the courtroom by providing more complete, reliable, and precise information for already well-established legal purposes. We just need to be careful not to overplay the neuroscientific evidence. We must also demand that when brain scans are introduced into the courtroom, they are obtained, analyzed, interpreted, and presented by qualified experts. In short, neuroscience is a vital tool in making decisions about competency, guilt, and sentencing, but it must be used properly to ensure fairness.

2.2 Neurobiologically Informed Risk Assessment

I turn now to a discussion of neurobiologically informed risk assessment methods. Decisions about public safety and the prevention of violence are more likely to succeed when they are based on accurate predictions of who will engage in violence and under what circumstances. The problem, however, is that accurately predicting violence before it occurs is not easy. As Eyal Aharoni et al. explain:

One potential way to reduce predictive error is to develop better risk assessment tools. Traditionally, offender risk estimates were determined using unstructured clinical judgment by trained forensic psychologists – a subjective technique that has since been shown to perform little better than chance. Recently, structured actuarial risk assessment techniques have improved the accuracy of offender placement decisions by quantifying the offender’s degree of fit with a known validation sample. As a result, more than twenty US states now require their courts to use these statistical tools in service of offender sentencing and placement decisions (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 161–62).

In the criminal justice context, structured actuarial risk assessment refers to any evidence-based technique that generates probabilistic predictions about the ostensible offender’s engaging in certain behavior (e.g., violence, reoffending, relapsing, or responding to treatment) by querying information about their attributes or circumstances (Reference TaxmanTaxman 2018; Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 161–62). Before we discuss the pros and cons of common risk assessment techniques, it is important to understand the variety of ways in which these techniques can be, and often are, employed in legal settings. Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. (2022: 163) provide the following very useful list:

• After a person is arrested, risk information might be useful to prosecutors making charging decisions. Prosecutors have wide discretion in such decisions. So, if a defendant is perceived to be particularly dangerous, a prosecutor might seek to file more charges or more severe charges.

• Bail decisions may also be informed by risk information.

• Risk assessments can also inform whether the individual is placed on probation or instead sentenced to prison. Risk scores are not typically used to influence the length of a prison sentence, but may be used in pre-sentencing reports and could be considered relevant to actual sentencing decisions—for example, in Texas, capital sentencing hinges in part on a determination of future dangerousness (Tex. Code Crim. Proc., Art. 37.071).

• Within a facility, risk information can determine security level assignments: Authorities may assign low, medium, or high supervision based on the offender’s recent behavior in custody, not as punishment but as a protective measure.

• A prison inmate could be released to parole prior to his sentence end date on grounds of a low risk assessment score.

• Treatment interventions can be based on risk assessment. In some states, such as Florida, judges can mandate treatments such as anti-androgen therapy (i.e., “chemical castration”) to certain sex offenders in addition to their original sentence based on their risk of reoffending (Fla. § 794.0235). Risk assessment can also inform decisions to administer treatment (typically noninvasive treatment) as a sentencing diversion.

In addition to these uses, risk assessment tools also lie at the heart of criminal justice reform to tackle mass incarceration. Melissa Hamilton, for instance, writes, “The newest application of risk tools centers on the pretrial stage as a means to reduce both reliance upon wealth-based bail systems and rates of pretrial detention” (Reference Hamilton2021a: 53).

This is an extremely important issue since three out of five people in US jails today have not been convicted of a crime. This amounts to nearly half a million people sitting in jail each day, despite being presumed innocent under the law. The vast majority of these individuals are awaiting trial but cannot afford the bail amount set for pretrial release. While the use of cash bail is ubiquitous in the United States, it is also fundamentally unjust (Reference CarusoCaruso 2020b). The cash bail system criminalizes poverty, as people who are unable to afford bail are detained while they await trial. In this way, cash bail perpetuates inequities in the justice system that are disproportionately felt by black and brown communities and those experiencing poverty. Furthermore, the legal rationale for pretrial detention is public safety. Yet the ability to pay cash bail is neither an indicator of a person’s guilt nor an indicator of risk in release. In fact, the cash bail system often leads to the detention of people who pose no real threat to public safety (Reference Laura and Arnold FoundationLaura and John Arnold Foundation 2013).

Despite its many potential advantages, critics fear that risk assessment tools may lead to negative consequences if they do not exhibit sufficient accurate predictions or do not treat protected groups fairly. One key concern is that current risk assessment tools tend to be algorithmic. They use big data and statistical analyses to derive correlates of criminal offending and package them into computational algorithms (Reference TaxmanTaxman 2018; Reference HamiltonHamilton 2021a, Reference Caruso2021b). If the data being used is already biased because of, say, the over policing of black and brown communities or historical biases elsewhere in the legal system, then the predictions produced by these tools will simply reflect and reinforce existing cultural biases. For this reason, the potential for race-based discrimination is an emerging concern regarding the use of algorithmic risk assessment methods (Reference HamiltonHamilton 2021a, Reference Caruso2021b). Critics fear that black defendants may be harmed when the algorithms learn on already biased data, imbed these structural inequalities, and further relaunder inequities through the guise of objective risk prediction.

This concern is a serious one, and we must do everything we can to guard against the possibility of algorithmic bias. That said, I also think it would be a mistake to completely abandon or prohibit the use of structured actuarial risk assessments over such concerns. For one, it’s possible that once we recognize the existence of such algorithmic bias, we can design better tools that avoid it. The first step is to recognize that such biases exist and that historic incarceration data may overestimate the risk of offending in certain populations. Once we recognize this, we can work to create more accurate and less biased tools for assessing risk – for example, tools that focus less on demographic factors, which may be more susceptible to bias, and more on factors unique to the individual, such as substance use problems, psychopathic personality features, anger, impulsivity, antisocial peers, antisocial attitudes, a history of violence, young age at the first violent act, stress, treatment nonadherence, lack of social support, and mental illness. Second, we have to weigh the use of structured actuarial risk assessment tools against the alternative – the status quo. I’ll argue in a moment that when we do this, we see that such tools, while not perfect, are a marked improvement over traditional methods of judging risk and dangerousness. Lastly, in the near future, neuropredictive technologies may allow us to further reduce bias by including unique neurobiological markers in our risk assessments.

Scientists, for instance, are beginning to examine “whether the inclusion of neurobiological and/or genetic markers can improve predictive models based solely on non-biological evidence” (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 168). In fact, a growing number of theorists have suggested that in the near future, neuropredictive technologies may be able to further improve our risk assessment methods, both in terms of accuracy and counterbalancing human bias (Reference Nadelhoffer, Bibas and GraftonNadelhoffer et al. 2012; Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022). Neuroscientists, for instance, have made tremendous progress in the past two decades in identifying and exploring some of the neural correlates of violence and aggression (Reference RaineRaine 2014).

Consider, as an example, what neuroscientists have learned about the brains of individuals with psychopathy, a developmental disorder that often leads to persistent antisocial behavior. Despite the fact that only 1 percent or less of the population is thought to be afflicted with psychopathy, some estimates suggest that individuals with psychopathy could nevertheless be responsible for as much as 30–40 percent of all violent crime (Reference Hare and McPhersonHare and McPherson 1984). Psychopaths are particularly prone to violence demonstrating increased aggressive behavior and committing a great number of violent attacks than non-psychopaths (Reference Salekin, Rogers and SewellSalekin et al. 1996). Psychopathy is also a strong predictor of how likely one is to re-offend after release from prison (Reference Porter, Birt and BoerPorter et al. 2001), and it is a particularly strong predictor of violent recidivism (Reference Cornell, Warren, Hawk, Staffoed, Oram and PineCornell et al. 1996; Reference Porter, Brinke and WilsonPorter et al. 2009). In fact, within one year of release psychopaths are about three times more likely to recidivate than non-psychopaths, and four times more likely to violently recidivate (Reference Hemphill, Hare and WongHemphill et al. 1998).

Over the last few decades, neuroscientists have begun to study the neuronal basis of psychopathy (see Reference Raine, Lencz, Bihrle, LaCasse and CollettiRaine et al. 2000; Reference Glenn, Raine, Yaralian and YangGlenn et al. 2010; Reference Glenn, Yang, Raine and SimpsonGlenn, Yang, and Raine 2012). Reference Leutgeb, Leitner and WabneggerLeutgeb et al. (2015), for example, compared structural imaging data from forty male high-risk violent offenders and thirty-seven non-delinquent healthy controls via voxel-based morphometry – a computational approach to neuroanatomy that measures differences in local concentrations of brain tissue. They then correlated psychopathic traits and risk for violence recidivism with gray matter volume of regions previously shown relevant for criminal behavior. They found that (a) relative to controls, criminals showed less gray matter volume in the prefrontal cortex and more gray matter volume in cerebellar regions and basal ganglia structures; (b) within criminals, there was a negative correlation between prefrontal gray matter volume and psychopathy; (c) there was a positive correlation between cerebellar gray matter volume and psychopathy as well as risk of recidivism for violence; (d) gray matter volumes of the basal ganglia and supplementary motor area were positively correlated with antisociality, and (e) that gray matter volume of the amygdala was negatively correlated with risk for violence recidivism. They concluded that in violent offenders, deviations in gray matter volume of the prefrontal cortex as well as areas involved in the motor component of impulse control (cerebellum, basal ganglia, supplementary motor area) are differentially related to psychopathic traits and the risk of violence recidivism. Other neuroimaging investigations have found reductions in orbitofrontal gray matter in psychopaths as well as volume reduction in the most anterior frontopolar regions of the prefrontal cortex (Reference De Oliveira-Souza, Hare and Bramatide Oliveira-Souza et al. 2008; Reference Tiihonen, Rossi and LaaksoTiihonen et al. 2008).

The amygdala also features prominently in theories of psychopathy due to its role in forming stimulus-reinforcement associations, conditioned fear responses, and the initiation of affective states. Recent neuroimaging data have strongly implicated the involvement of the amygdala in psychopathy-related deficits (Reference Anderson and KiehlAnderson and Kiehl 2014). In one large-scale investigation involving nearly 300 incarcerated subjects, Reference Ermer, Cope, Nyalakanti, Calhoun and KiehlErmer et al. (2011) found that psychopathy was associated with decreased regional gray matter in several paralimbic and limbic areas, including the amygdala. Reference Yang, Wong and CoidYang et al. (2010) also found that volume reductions in both the prefrontal cortex and the amygdala were more pronounced in psychopaths with criminal convictions compared to both controls and “successful” psychopaths. Given these new insights into the neurological correlates of psychopathy, neuroscientific methods may have the potential to improve existing tools for predicting violent recidivism (Reference Leutgeb, Leitner and WabneggerLeutgeb et al. 2015: 194).

Beyond psychopathy, it’s also possible that we may be able to improve predictions of antisocial behavior more generally by including targeted brain metrics in risk assessment models. For instance, in a series of studies conducted by Reference Aharoni, Vincent and HarenskiAharoni et al. (2013), they found that in a sample of ninety-six adult offenders who engaged in an impulse control task while undergoing functional magnetic resonance imaging, brain activity within the anterior cingulate cortex (ACC) prospectively predicted being rearrested later. Offenders with relatively low ACC activity had roughly double the odds of getting rearrested for a violent or nonviolent crime as those with high activity, controlling for other known risk factors. In fact, using advanced statistical techniques, these neuropredictive models demonstrated relatively high accuracy – that is, the probability that a true positive was correctly classified exceeded 75 percent (Reference Aharoni, Mallett and VincentAharoni et al. 2014). Other studies have reported similar predictive effects using different models of brain function (Reference Kiehl, Anderson and AharoniKiehl et al. 2018; Reference Delfin, Krona and AndinéDelfin et al. 2019). Together, “these studies provide preliminary evidence that brain factors such as ACC function may serve as candidate biomarkers for antisocial behavior” (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 170). Perhaps in the future, then, we will be able to additionally improve existing actuarial risk assessment tools by incorporating neuropredictive technologies, further increasing their accuracy and reducing subjective bias.

Despite these advances in actuarial risk assessment, its use in legal settings continues to be the subject of much scholarly debate. Many ethicists, for instance, suggest that “all actuarial risk assessment is too problematic for use in justice settings, cautioning about violations of beneficence (e.g., unjustified harm to offenders), justice (e.g., unfair distribution of sanctions), and respect for persons (e.g., unjustified restrictions on the offender’s freedom or exposure to his private mental life), among other problems” (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 162). Such opposition, however, needs to be considered in contrast with its alternative: the status quo. Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. (2022) examine such ethical concerns and argue that:

I agree with Aharoni et al. that, when applied to various uses by the law, actuarial risk assessment often fares better on key ethical criteria than traditional clinical methods.

One common concern critics raise is that the predictive accuracy of actuarial risk assessments is insufficient for widespread use. This concern, however, is misleading since meta-analysis has demonstrated that the predictive value of many actuarial risk assessment instruments is superior to the traditional methods that would otherwise be used (Reference Grove and MeehlGrove and Meehl 1996). Traditional risk assessment techniques rely heavily on unstructured clinical judgments to assess the risk of antisocial behavior. Given that clinical assessment is inherently subjective, “it is unsurprising that the resulting predictions of future dangerousness are so unreliable” (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 164). In fact, “some commentators have even gone so far as to suggest that relying on clinical risk assessment for the purposes of the law is tantamount to ‘flipping coins in the courtroom’” (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 164). Worse still, “the unguided and intuitive nature of the process also makes it possible for biases and prejudices of clinicians to influence their assessments.” It’s my view, then, that while we should continue to strive to increase the power of actuarial risk assessment, it already represents a significant step forward, “even if it falls short of some critics’ idealized standards” (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 171).

Another main class of concerns critics raise has to do with potential violations of norms of beneficence, justice, and respect for persons. For example, “false-positive errors could support a longer sentence or revocation of privileges from a non-dangerous offender, and false-negative errors could result in an early release of a dangerous offender, placing the community members at risk of victimization” (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 173). To avoid these errors, one may be tempted to prohibit the use of actuarial risk assessment in an effort to minimize unnecessary harm. Unfortunately:

The central point here is that if we remove all actuarial risk assessments, violations of beneficence, justice, and respect for persons will not simply go away. Instead, they may actually increase since the alternative yields even greater classification errors.

The question, then, should not be whether the use of risk judgments is without serious and important moral concerns, but whether the use of particular actuarial risk factors is any more problematic than the other alternatives (Reference Nadelhoffer, Bibas and GraftonNadelhoffer et al. 2012; Reference Douglas, Pugh, Singh, Savulescu and FazelDouglas et al. 2017; Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022). I contend that in many cases it may actually be less problematic. This is not to advocate, of course, for the uncritical use of actuarial risk assessment. Instead, the justification of such assessments must be judged relative to the ways in which they are to be used and what the other alternative options are. Of course, for these tools to be effective and ethical, developers need to continue to work on minimizing bias and increasing accuracy. But here too actuarial risk assessment may have an advantage over traditional methods: “By design, actuarial risk assessment attempts to codify risk factors. This feature makes the process transparent and enables us to monitor and evaluate procedural mistakes and ultimately to minimize bias” (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 178). For this reason, risk assessment tools enable the assumptions of the predictive models to be explicit and subject to evaluation and revision. This level of transparency is not found in traditional clinical methods. As a result, if we dispense with structured actuarial risk assessment in justice decisions, “assessment bias does not go away, it just exerts its nasty effects behind a curtain of opacity” (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 179).

Moving forward, then, I suggest that the results of independent research can productively inform policymakers and stakeholders who are interested in ensuring the utility and fairness of the assessment instruments they may employ. The use of risk assessment is ubiquitous in our legal system. Since the use of risk assessment is not likely going away, we should continue to study and evaluate different actuarial risk assessment instruments, adopting those with the highest rate of predictive accuracy and the lowest rate of bias. We should also continue to work on ways to improve our existing assessment methods, which could include adding neurobiological measures to characterize biological markers of human behavior that increase the ability to predict particular behavioral outcomes accurately. A good test for whether we should do this would be whether explanations of human behaviors demonstrate greater predictive accuracy when the relevant biological, neural, psychological, and social factors are all included in the model (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022: 169).

2.3 Mind Reading and Lie Detection

I turn now to another controversial topic in the domain of assessment: the use of neuroimaging techniques in “mind reading” and lie detection. Traditional lie detection tools, such as polygraph, voice stress analysis, or special interrogation techniques, rely on behavior or psychophysiological manifestations of deception. With the advent of neuroimaging techniques, “the question emerged whether it would be possible to directly identify deceit in the part of the body where it is generated: the brain” (Reference GamerGamer 2014: 172). After a few promising studies, these techniques became commercially available, and there have been attempts to use such results in court – though such attempts have thus far been unsuccessful. In this section, I will briefly review the development of neuroimaging techniques in the field of deception detection and critically discuss the potential benefits and shortcomings of such methods. I will also discuss some of the moral and legal issues associated with such technologies, such as concerns over rights and privacy.

Within the law, a reliable and valid detection of deception is of significant interest and has attracted a substantial amount of research during the last few decades. As Matthias Gamer writes, “From an applied perspective, such [a] method should be easy to handle, applicable to individual cases, resistant to countermeasures, and yield accurate diagnoses” (Reference Gamer2014: 172). When polygraphy was first introduced, it promised to provide just such a method and was widely adopted by investigative authorities all over the world. A polygraph is a device or procedure that measures and records several physiological indicators such as blood pressure, pulse, respiration, and skin conductivity while a person is asked and answers a series of questions. The theory underpinning the use of the polygraph is that deceptive answers will produce physiological responses that can be differentiated from those associated with nondeceptive answers. However, while polygraphs are sometimes used as an interrogation tool with criminal suspects or candidates for sensitive public or private sector employment, they have questionable reliability and, as a result, are generally not admissible as evidence in court. In fact, assessments of polygraphy by scientific and government bodies generally suggest that polygraphs are highly inaccurate, may easily be defeated by countermeasures, and are an imperfect or invalid means of assessing truthfulness (US Congress Office of Technology Assessment 1983; American Psychological Association 2004; Reference GamerGamer 2014).

One promising new method makes use of neuroimaging techniques, which have opened the way for a new and different approach to detecting deception. The key idea is that by examining the brain directly, we can hopefully find “a specific signature of deception in the brain activity that would allow for a sensitive and accurate detection of deceit” (Reference GamerGamer 2014: 173). One advantage of using neuroimaging techniques is that brain activity can be noninvasively measured in humans, either by recording cortical electric activity from the scalp using electroencephalography or by measuring neurovascular changes with Positron Emission Tomography (PET) or functional magnetic resonance imaging (fMRI).

One of the first studies to use functional magnetic resonance imaging (fMRI) to detect deception at the individual level was conducted by Reference Kozel, Johnson and QiwenKozel and colleagues (2005). In the study, subjects participated in a mock crime and were instructed to either steal a ring or a watch from a drawer. Kozel et al. then used fMRI to show that specific regions were reproducibly activated when subjects deceived. While undergoing fMRI, subjects were asked questions about both objects (e.g., “Did you steal the watch?” “Did you take the ring from the drawer?”) and instructed to deny stealing anything. As participants were familiar with both objects and delivered the same response (“No”), these questions only differ with respect to truthful or deceptive responding. In the study, a Model-Building Group (MBG, n = 30) was used to develop the analysis methods, and the methods were subsequently applied to an independent Model-Testing Group (MTG, n = 31). Reference Kozel, Johnson and QiwenKozel et al. (2005) were able to correctly differentiate truthful from deceptive responses, correctly identifying the object stolen, for 93 percent of the subjects in the MBG and 90 percent of the subjects in the MTG.

Comparable results were observed in other studies that used different statistical procedures (see, e.g., Reference Davatzikos, Ruparel and FanDavatzikos et al. 2005; Reference Ganis, Kosslyn, Stose, Thompson and Yurgelum-ToddGanis et al. 2011). For instance, Reference Davatzikos, Ruparel and FanDavatzikos et al. (2005) used a multivariate nonlinear high-dimensionality pattern classification method applied to fMRI images to discriminate between the spatial patterns of brain activity associated with lying and truth. In the study, twenty-two participants were presented with an envelope containing two cards (five clubs and seven spades). In the imaging phase of the study, the investigator instructed participants to deny possession of one of the cards and acknowledge possession of the other. Davatzikos and colleagues were able to correctly discriminate 99 percent of the true and false responses. These results, Davatzikos et al. conclude, “demonstrate the potential of non-linear machine learning techniques in lie detection and other possible clinical applications of fMRI in individual subjects, and indicate the accurate clinical tests could be based on measurements of brain function with fMRI” (Reference Davatzikos, Ruparel and Fan2005: 663).

While these results look promising, we are still a long way off from reliable neuroimaging-based lie detection. For a deception detection method to be trustworthy, not only must it be applicable to individual cases and yield accurate diagnoses, but it must also be resistant to countermeasures. However, Reference Ganis, Kosslyn, Stose, Thompson and Yurgelum-ToddGanis et al. (2011) tested whether countermeasures – methods prevaricators employ to confuse deception detection procedures – could defeat fMRI deception tests. Using a modified concealed information test (CIT) in which participants lied about knowing their birth date when shown six dates on a computer, Reference Ganis, Kosslyn, Stose, Thompson and Yurgelum-ToddGanis et al. (2011) instructed some participants how to perform a countermeasure consisting in associating distinct covert actions to the irrelevant dates in the sequence. When compared to those participants who did not receive training in countermeasures, Ganis et al. found that deception detection accuracy fell to only 33 percent in participants with countermeasures training. As a result, they conclude: “These findings show that fMRI-based deception detection methods can be vulnerable to countermeasures, calling for caution before applying these methods to real-world situations” (Reference Ganis, Kosslyn, Stose, Thompson and Yurgelum-ToddGanis et al. 2011: 312).

A second critical issue discussed in the literature has to do with the fact that the vast majority of studies in this domain use artificial settings that guarantee experimental control at the expense of real-world validity (Reference Greely and IllesGreely and Illes 2007; Reference GamerGamer 2014). As Gamer describes, “with the exception of very few studies …, participants were always instructed to lie to specific questions and were not free to decide about their behavior during the interrogation procedure” (Reference Gamer2014: 178). Moreover, “participants had nothing to lose when failing the test which sharply contrasts with the situation in the field where test results might have devastating consequences for the examinee.” It remains an open question, then, how neuroimaging techniques will fare in real-world situations where the stakes are high and laboratory controls are not possible.

A third concern has to do with the lack of diversity of subjects in these studies (Reference Greely and IllesGreely and Illes 2007; Reference Rusconi and Mitchener-NissenRusconi and Mitchener-Nissen 2013). For the most part, these experiments used healthy young adults with little gender or ethnic diversity. “No one tested children, the middle-aged or elderly, those with physical or mental illnesses, or those taking drugs, either as medication or illicitly” (Reference Greely and IllesGreely and Illes 2007: 403). This is problematic since the criminal justice system interacts with many types of individuals that are not taken into account in these fMRI studies. For neuroimaging lie detection to be reliable in the real world, it needs to be “able to cope with the types of individuals usually encountered by law enforcement officers, including substance addicts, those with high incentives to lie, and those with mental disorders” (Reference Rusconi and Mitchener-NissenRusconi and Mitchener-Nissen 2013: 6). The lack of extensive research on such populations is a huge drawback. In fact, doubts already exist as to whether fMRI would be usable for those presenting with conditions such as delusions and amnestic disorders with confabulation (Reference Langleben, Dattilo and GutheiLangleben et al. 2006).

A fourth concern has to do with the inconsistency of reported brain regions of activity in these studies. As Greely and Illes write: “[A]ll of the relevant experiments report finding activation of various regions of the brain (sometimes defined narrowly, sometimes broadly). Together, they find activation in many different areas of the brain without strong consistency among the experiments, except when brain regions are very broadly defined” (Reference Greely and Illes2007: 403). This diversity “casts some doubt on the accuracy of any particular method of lie detection” (Reference Greely and IllesGreely and Illes 2007: 403). The problem is that researchers cannot agree on exactly which brain regions denote deception, at least not with any degree of precision.

A closely related concern is that, given the diversity of brain regions identified in these studies, how are we to demonstrate which, if any, constitute the neural correlates of lying? Perhaps some of these regions are involved in subjects exerting extra effort, rather than being the neural correlate of the lie itself. After all, lying often requires extra effort compared to responding truthfully. Such extra effort may be aimed at inhibiting the truth and/or producing an alternative response that sounds realistic. Given the very real possibility, then, that some of these regions may be associated with tasks related to or involved in lying, but not lying itself, it would be very difficult to prove that a person is lying whenever a particular region activates during a task. In fact, “many defendants [while testifying] do inhibit their natural tendency to blurt out everything they know … Many of them also suppress expression of anger and outrage at accusations. Suppressing natural tendencies is not a reliable indicator of lying, in the context of trial” (Reference Grafton, Sinnott-Armstrong, Gazzaniga and GazzanigaGrafton et al. 2006: 36–37). Given that fMRIs merely detect and measure manifestations of thoughts through changes in oxygenated blood, which proponents consider denotes lying, “any exhibited increase in blood flow detected by fMRI may result from alternative neurological processes such as anxiety, fear, or other heightened emotional states which are unrelated to the question of deception” (Reference Rusconi and Mitchener-NissenRusconi and Mitchener-Nissen 2013: 5). In other words, just because certain brain regions are activated during deception, it does not follow that every time those regions are activated the individual is lying.

Beyond these rather technical concerns, there are also important legal and ethical concerns that need to be considered. As Rusconi and Mitchener-Nissen write:

Among the legal and ethical concerns raised by neuroimaging technologies are possible constitutional and human rights violations regarding illegal search, right to silence, freedom of thought, right to privacy, human dignity, right to integrity of the person, and protection of personal data (see Reference Rusconi and Mitchener-NissenRusconi and Mitchener-Nissen 2013; Reference Greely and IllesGreely and Illes 2007; Reference FarahanyFarahany 2023).

While I cannot discuss all these issues here, consider the question of whether fMRI-based lie detection constitutes a search of the subject, and if so, under what conditions such a search will be considered lawful or unlawful. Discussions in this area tend to center on the US Constitutional Fourth Amendment, which protects against unreasonable or unlawful searches. According to some commentators, neuroimaging techniques will constitute a legitimate search under established legal doctrine should neural activity be equated to other forms of physical evidence gathered from the human body, such as blood or DNA sampling, fingerprints, voice samples, etc. – provided that probable cause exists justifying such sampling. However, as Rusconi and Mitchener-Nissen note: “it is easy to conceptualize neural activity as distinct from other forms of physiological evidence” (Reference Rusconi and Mitchener-Nissen2013: 8). For example, “while we can manipulate neural activity by conducting mathematical problems in our head, we cannot change our DNA profile through thought processes” (Reference Rusconi and Mitchener-Nissen2013: 8). What legal weight such a distinction should carry has yet to be determined, but neurolaw practitioners need to be addressing these issues now.

Another question that needs to be considered is whether or not authorities should be allowed to record our neural activity without our consent. If it is determined, for instance, that neuroimaging-based lie detection is a search subject to Fourth Amendment regulation, government authorities can perform it as long as they obtain a warrant. But should the state be allowed to access private mental activity without consent? On the one hand, the courts may rule that neural activity is unlike other kinds of evidence and that neural-searches are a violation of individuals’ privacy. On the other hand, the US Supreme Court has already ruled that in matters of public safety, some violations of privacy may be justified (Katz v. United States 389 US 347 [1967]). For instance, a defendant may sometimes be required to undergo diagnostic tests and even medical surgeries in order to seize potentially probative evidence such as a bullet or ingested jewelry (Reference Aharoni, Abdulla, Allen, Nagelhoffer, Brigard and Sinnott-ArmstrongAharoni et al. 2022). When confronted with this problem, courts will be forced to either shoehorn this new technology into existing legal frameworks governing conceptually similar subject matter (i.e., DNA, blood, fingerprints, etc.) or produce new bespoke legal frameworks for this governance (Reference Rusconi and Mitchener-NissenRusconi and Mitchener-Nissen 2013: 8).

Another potential legal concern is whether fMRI questioning undermines the right to silence and the right not to self-incriminate. Rusconi and Mitchener-Nissen sketch the problem as follows:

It’s easy to see, then, that important moral and legal questions remain unanswered – not only with regard to the right to silence and what constitutes lawful search, but also with regard to other rights as well.

Because of concerns like these, some commentators have argued that what is needed is the recognition of a new set of “neurorights” (see Reference McCayMcCay 2022; Reference FarahanyFarahany 2023). Advocates of neurorights maintain that we “must establish the right to cognitive liberty – to protect freedom of thought and rumination, mental privacy, and self-determination over our brains and mental experiences” (Reference FarahanyFarahany 2023: 11). But while the case for neurorights is sometimes made in light of neuroimaging-based lie detection, it is more often directed toward the use of neurotechnologies more generally (Reference McCayMcCay 2022; Reference FarahanyFarahany 2023). Neurotechnologies are devices that monitor the brain or nervous systems of people, and may act on them to influence neural activity (Reference McCayMcCay 2022). Sometimes neurotechnologies are implanted in the brain, or interact with the nervous system through a headset or another “wearable” device. They are used in research and in therapeutic contexts, in computer gaming, meditation, and sometimes in the workplace to monitor attention. The concern is that large amounts of “brain data” collected by neurotech devices could allow people, governments, and corporations with access to hack our brains, make inferences about our mental states, and potentially profit off of us. This raises concerns about mental privacy and commercial manipulation. It’s important, therefore, that before we permit the use of neuroimaging-based lie detection in criminal courts, we first consider whether new neurorights protections are needed or whether existing rights and protections can sufficiently address the concerns just raised.

To summarize, while initial studies suggest that neuroimaging-based lie detection may one day become a viable alternative to current methods, we have a long way to go before its reliability and accuracy are proven effective enough for use in the real-world. Current fMRI-based studies on deception detection have shown promising results, but they are still vulnerable to countermeasures, use highly artificial settings, and lack the kind of diversity of subjects likely to be encountered within the criminal justice system. It’s also unclear whether the kinds of lies tested in these studies resemble the full diversity of situations, settings, and kinds of lies found in the real world. Concerns also remain about the inconsistency of reported brain regions identified in these studies, as well as the difficulty of determining which regions constitute the neural correlates of lying – and not, say, regions associated with tasks related to, or involved in, lying. Beyond these concerns, there are also important ethical and legal issues that still need to be addressed. If neuroimaging methods were one day perfected, it does not automatically follow that we should permit their use in court. Accuracy and reliability are not the only issues. We also need to determine whether their use violates important human and constitutional rights.

3.1 The Benefits of Neurointerventions

To begin, we should note that there are several potential benefits associated with the use of neurointerventions. First and foremost, neurointerventions have the potential to reduce crime, suffering, and financial and personal costs. Imprisonment comes at a high economic cost to society. The Bureau of Justice Statistics estimates that the United States alone spends more than $80 billion each year to keep roughly 2.3 million people behind bars – and many experts say that figure is a gross underestimate. Incarceration also comes at great personal cost to prisoners – including separation from family and friends, loss of liberty, and stigmatization. Hence, “if neurointerventions are developed to effectively aid criminal rehabilitation, then these interventions have the potential to generate economic savings and to spare offenders suffering by allowing them to avoid incarceration or be released earlier” (Reference ShawShaw 2022: 1421; see also Reference ShawShaw 2018a).

Neurointerventions also have the potential to assist in the treatment and rehabilitation of individuals involved in the criminal justice system. By addressing underlying neurological issues or enhancing cognitive functioning, these interventions potentially contribute to reducing the likelihood of recidivism. The potential of neurointerventions also extends beyond rehabilitation to the prevention of crime. By identifying individuals at risk through neurobiological markers, early intervention with targeted neurointerventions can disrupt the trajectory towards criminal behavior. This proactive approach, based on the underlying neurological factors, may contribute to crime prevention.

One of the most widely discussed and debated neurointerventions is “chemical castration,” which is a medical intervention that involves the use of anti-libidinal medications to suppress or reduce sexual drive and functioning (Reference Douglas, Bonte, Focquart, Devolder and SterckxDouglas et al. 2013; Reference DouglasDouglas 2014a; Reference Forsberg and DouglasForsberg and Douglas 2017; Reference Sifferd, Nadelhoffer and VincentSifferd 2020; Reference ShawShaw 2022). It is primarily used as a treatment for individuals who have committed sexual offenses, particularly those involving pedophilia or other forms of sexual violence. The term “castration” is used metaphorically here since the medications used do not physically remove or alter the genital organs – and unlike physical castration, chemical castration is generally reversible when treatment is discontinued (although it can produce permanent side effects). The medications used in chemical castration typically fall into two categories: anti-androgens and gonadotropin-releasing hormone (GnRH) agonists. Anti-androgens work by blocking the effects of androgens, such as testosterone, in the body. GnRH agonists, on the other hand, inhibit the release of gonadotropins, which are hormones that stimulate the production of testosterone or estrogen.

In the literature on neurointerventions, “anti-libidinal drugs that reduce the effects or production of testosterone are considered neurointerventions because they have a considerable impact on the brain and on aspects of the individual’s psychology such as thoughts, desires, and motivations” (Reference ShawShaw 2022: 1414). This is important since the effects of anti-libidinal drugs on the brain are more important in addressing sexual offending than the drugs’ physical effects on sexual function (Reference Greely and FarahanyGreely and Farahany 2019). As Shaw correctly notes, “Even if someone’s ability to engage in penetrative sexual activity has been reduced/removed, the person can still find ways of committing serious sexual offenses if the desire/motivation to do so remains” (Reference Shaw2022: 1414–15). In theory, then, anti-libidinal drugs function as a neurointervention by moderating the activity of testosterone, which has effects on the responsiveness of both general and specific neurological arousal mechanisms, influences the processing of sexual sensory stimuli, and impacts the motivation, attention, and mood associated with aggression and dominance (Reference Grubin, Beech, Carter, Mann and RothsteinGrubin 2018: 711–12).

Currently, a variety of jurisdictions administer anti-libidinal drugs to sex offenders “under either statutory provisions that specifically concern sexual behavior or general mental health legislation” (Reference ShawShaw 2022: 1415). Depending on the jurisdiction, “these interventions may be offered on a voluntary basis within prison or in the community, imposed as a compulsory treatment in the community or as a mandatory condition of parole, or exchanged for early release” (Reference ShawShaw 2022: 1415). For instance, several European countries allow for only the voluntary use of chemical castration, which means that it cannot be imposed as a compulsory penalty but can be offered as a treatment option or an alternative to further incarceration. In Poland, however, it is possible to sentence offenders to compulsory chemical castration. In 1996, California law was amended to require chemical castration for repeat child molesters upon parole (California Penal Code, Sect. 645).

While the use of chemical castration is controversial for a number of reasons, including concerns over consent, coercion, effectiveness, and side effects, I mention it here only to highlight one widely discussed neurointervention. In what follows, I’ll discuss all the various concerns that arise with this and other forms of neurointervention, especially when they are mandated by the state rather than offered voluntarily. It’s important to recognize, however, that neurointerventions are already being utilized within the criminal justice system to rehabilitate offenders and reduce recidivism. And while some neurointerventions are rather invasive and/or come with significant risks or side effects, others are rather benign and have very few, if any, side effects (Reference Choy, Focquaert and RaineChoy, Focquaert, and Raine 2020).

For instance, one rather benign form of intervention is the use of neurofeedback therapy in correctional settings. Neurofeedback is “an innovative approach that may ultimately lessen criminal behavior, prevent violence, and lower recidivism” (Reference Gkotsi and BenaroyoGkotsi and Benaroyo 2012: 3). As Gkotsi and Benaroyo describe:

Reference QuirkDouglas Quirk (1995), a Canadian researcher, tested the effects of a neurofeedback treatment program on seventy-seven dangerous offenders in an Ontario correctional institute who suffered from deep-brain epileptic activity. The results demonstrated a reduction in the subjects’ criminal recidivism and suggested that “a subgroup of dangerous offenders can be identified, understood, and successfully treated using this kind of biofeedback conditioning program” (Reference Quirk1995). Additional studies on juvenile offenders with significant psychopathology and electroencephalographic abnormalities (Reference Smith and SamsSmith and Sams 2005), and on male adolescents diagnosed with ADHD (Reference Martin and JohnsonMartin and Johnson 2005), also demonstrated reduced recidivism, improved cognitive performance, improved emotional and behavioral reactions, and inhibition of inappropriate responses. These results are promising and suggest that neurofeedback could potentially provide a noninvasive, cost-effective method for reducing recidivism.

Additional noninvasive forms of neurointerventions include transcranial magnetic stimulation, omega-3 supplementation, and mindfulness and meditation techniques. Something as benign as omega-3 supplements, for instance, has been proposed as a way to reduce aggression and antisocial behavior in children and adults (Reference Raine, Portnoy, Liu, Mahoomed and HibbenRaine et al. 2015; Reference Choy, Focquaert and RaineChoy, Focquaert, and Raine 2020) and studies have been conducted on its ability to reduce aggressive behavior in prison populations (Reference Meyer, Byrne and CollierMeyer et al. 2015; 202). Transcranial Magnetic Stimulation (TMS) and Transcranial Direct Current Stimulation (tDCS) are two additional promising noninvasive neurointerventions. TMS is a noninvasive form of brain stimulation that involves the use of magnetic fields to stimulate or inhibit specific brain regions through electromagnetic induction. During a TMS session, an electromagnetic coil is placed against the scalp, which creates a varying magnetic field, inducing a current within a region in the brain itself. It can be employed to target areas associated with impulse control, aggression, or emotional regulation, with the aim of reducing criminal behavior. tDCS, on the other hand, is used to modulate cortical excitability, producing facilitatory or inhibitory effects upon a variety of behaviors. The main differences between the two are that TMS uses repetitive electromagnetic pulses while tDCS uses a constant weak electrical current, TMS is a neuro-stimulator while tDCS is a neuro-modulator, and tDCS equipment is typically small, battery-powered, and portable, while TMS devices are wall-powered, larger, and heavier.

While the effectiveness of these neurointerventions in reducing criminal behavior is still an area of ongoing research, several studies have shown good results in addressing various behaviors associated with criminal conduct, such as aggression, anger, irritability, and impulsivity (see Reference Romero-Martinez, Bressanutti and Moya-AlbiolRomero-Martinez, Bressanutti, and Moya-Albiol 2020; Reference Philipp-Wiegmann, Rosler and RomerPhilipp-Wiegmann et al. 2011; Reference Choy, Raine and HamiltonChoy, Raine, and Hamilton 2018). For instance, in one study using transcranial direct current stimulation, stimulation of the dorsal anterior cingulate cortex resulted in improved performance on cognitive and affective attentional tasks (Reference To, Eroh, Hart and VannesteTo et al. 2018). In another study using the same technology, a single stimulation session of the dorsolateral prefrontal cortex reduced the desire to carry out violent antisocial acts by over 50 percent (Reference Choy, Raine and HamiltonChoy, Raine, and Hamilton 2018). Studies by Reference Fecteau, Knoch and FregniFecteau et al. (2007) and Reference Boggio, Villani and ZaghiBoggio et al. (2010) also found that stimulation of the left dorsolateral prefrontal cortex produced a decrease in risk-taking, a behavior associated with increased inhibitory control.

We can see, then, that neurointerventions have the potential to contribute to crime prevention in several ways. Some potential benefits include rehabilitation and recidivism reduction. By combining neurointerventions with therapy, counseling, or behavioral interventions, it may be possible to enhance treatment outcomes and reduce the likelihood of recidivism. Neurointerventions can also target and address underlying neurological issues that contribute to criminal behavior. By modulating brain activity or addressing imbalances, these interventions may help reduce impulsivity, aggression, or other behaviors linked to criminal conduct. Certain neurointerventions, such as cognitive training or neurofeedback, can also improve cognitive functioning and enhance self-regulation skills. These improvements can support individuals in making better decisions, managing emotions, and developing the necessary skills to avoid engaging in criminal activities. Finally, neurointerventions have the potential to provide personalized treatment plans based on an individual’s neurobiological profile. This tailored approach can address specific cognitive or behavioral deficits that may contribute to criminal behavior, leading to more effective interventions.

3.2 Safety and Effectiveness

While the use of neurointerventions offers a number of promising benefits, it’s important to note that addressing and reducing crime involves a complex interplay of various factors, including social, environmental, psychological, and economic aspects. Hence, the reduction of crime requires comprehensive approaches that involve social interventions, community support, evidence-based rehabilitative programs, and addressing underlying causes such as poverty, substance abuse, homelessness, and mental health issues (Reference CarusoCaruso 2021a). While neurointerventions may have a role to play as part of a broader approach to crime, they alone do not constitute a comprehensive solution to crime reduction. Furthermore, the use of neurointerventions should be approached with caution since there are important ethical considerations that must first be addressed, including concerns over safety, consent and coercion, individual rights, and personal identity.

Reference GreelyHenry Greely (2007), for instance, has argued that the use of neurointerventions within the criminal justice system is only acceptable if the interventions are safe and effective. Though this seems simple enough, we need to keep in mind that “when it comes to interventions against criminal behavior, proving safety and efficacy is fraught with difficulties” (Reference Greely2007: 1116). As Choy, Focquaert, and Raine note, “for many interventions, establishing adequate safety and effectiveness is challenging given that clinical trials in offender populations face serious hurdles due to practical, ethical, and regulatory limitations” (Reference Focquaert, Van Assche, Sterckx, Vincent, Nadelhoffer and McCay2020: 5). For instance, in many cases, psychopharmacological interventions “are known to be used off-label [not officially approved] in offender populations without guarantees concerning effectiveness in reducing recidivism or adequate implementation limitation to prevent irreversible side effects” (Reference Choy, Focquaert and RaineChoy, Focquaert, and Raine 2020: 33). To make things even more complicated, a significant number of offenders suffer from mental disorders and therefore face “double vulnerability.”

If neurointerventions are to be used ethically, adequate safeguards are essential. One proposal, which I find reasonable, demands, minimally, that the treatment or neurointervention in question “must be considered standard medical therapy before its use is considered within the criminal justice system” (Reference Choy, Focquaert and RaineChoy, Focquaert, and Raine 2020: 33). This, at least, prohibits the use of experimental neurointerventions on offenders within the criminal justice system or off-label use of psychopharmacological interventions. I favor this approach over the more extreme option, which is to prohibit the use of all neurointerventions. That’s because, if benign and effective biological interventions exist, not allowing offenders to benefit from these treatments would make an already vulnerable group even more vulnerable by excluding the possibility of a future crime-free life. Furthermore, as Neil Levy’s parity principle stipulates: “Unless we can identify ethically relevant differences between internal and external interventions and alterations, we ought to treat them on a par” (Reference Levy2007: 62).

Of course, important questions still remain, such as how safe and effective must an intervention be proven before we consider it ethically acceptable? While I do not propose to answer this question here, I maintain that the same ethical considerations that guide the use of medical, biological, and psychopharmacological interventions in general should guide the use of neurointerventions within the criminal justice system more specifically. These include the medical ethical principles of beneficence and nonmaleficence. The principle of beneficence requires a moral obligation to act for the benefit of others. The principle of nonmaleficence, on the other hand, holds that there is an obligation not to inflict harm on others and is closely associated with the maxim primum non nocere (first do no harm). Since many treatments involve some degree of harm (e.g., side effects from drugs, chemotherapy, etc.), these principles are typically interpreted as implying that the harm should not be disproportionate to the reasonably expected benefits of treatment. While there are examples of neurointerventions that clearly run afoul of these principles and therefore should be prohibited, there will also be safe and effective neurointerventions that satisfy these principles.

For instance, pharmacological interventions for drug or alcohol addiction, in conjunction with other anti-addiction counseling and treatments, often produce, on the whole, a balance of benefits over risks/harms and can be quite effective in combating addiction. When this is the case, and these treatment programs use FDA-approved drugs to combat drug and alcohol addiction, we can say that they satisfy the principles of beneficence and nonmaleficence and are prima facie ethically permissible. Addiction to heroin or other opioids, for instance, can be treated with methadone, which prevents withdrawal symptoms and reduces the craving for heroin, while avoiding many of the risks and harms of heroin. In fact, abundant evidence shows that methadone, buprenorphine, and naltrexone all reduce opioid use and opioid use disorder-related symptoms, and that they reduce the risk of infectious disease transmission as well as criminal behavior associated with drug use (American Society for Addiction Medicine 2017; National Institute of Drug Abuse 2021). These medications also increase the likelihood that a person will remain in treatment, which itself is associated with a lower risk of overdose mortality, reduced risk of HIV and HCV transmission, reduced criminal justice involvement, and a greater likelihood of employment. While these drugs are not without side effects and risks, it is often thought that their benefits outweigh the risks. When this is the case, and these drugs are considered standard medical therapy and approved by the FDA, we may object to their use within the criminal justice system on other ethical grounds (e.g., consent and coercion), but not on the grounds of safety and effectiveness. It’s the job of experts, therefore, to determine which types of interventions are safe and effective, when the benefits outweigh the costs, and when these interventions can be employed within the criminal justice system without exploiting vulnerable populations or violating other important ethical considerations.

3.3 Consent and Coercion

In addition to concerns over safety and effectiveness, the use of neurointerventions also raises complex questions about consent and coercion – especially among imprisoned populations. Consider, for instance, the principle of autonomy, another common medical ethical principle. In medical practice, autonomy is usually expressed as the right of competent adults to make informed decisions about their own medical care. The principle underlies the requirement of informed consent of patients before any investigation or treatment takes place. Two conditions are ordinarily required before a decision can be regarded as autonomous. The individual has to have the relevant internal capacities for self-government and has to be free from external constraints. Does the use of neurointerventions within the criminal justice system run afoul of the principle of autonomy? And, if so, is this ever justified?

To begin, I would first like to note that some theorists have argued that offering neurointerventions under the right circumstances can actually increase the autonomy and well-being of offenders (Reference Douglas, Bonte, Focquart, Devolder and SterckxDouglas et al. 2013; Reference FocquaertFocquaert 2014; Reference RybergRyberg 2019; Reference Choy, Focquaert and RaineChoy, Focquaert, and Raine 2020). As Choy and colleagues explain:

Douglas et al., for instance, have argued that chemical castration can “increase future autonomy overall, either by removing internal barriers (such as irrational, inauthentic, compulsive desires) or external ones (such as restrictions on free movement)” (Reference Douglas, Bonte, Focquart, Devolder and Sterckx2013: 399). The thought here is that desires that drive many sex offenders frequently constitute very severe impediments to autonomy and are often at odds with the individual’s own higher-order desires (not to offend, to remain ethical, and to resist their compulsive sexual desires). By removing (or reducing) these impediments, the individual’s autonomy and authentic-self can be restored.

Derk Pereboom and I have made a similar point with regard to reasons-responsiveness (Reference Pereboom, Caruso, Caruso and FlanaganPereboom and Caruso 2018). We maintain that methods of therapy that engage reasons-responsive abilities should always be preferred – for example, consoling, talk therapies, and educational programs. Still, a concern for many forms of therapy proposed for altering criminal tendencies is that they circumvent, rather than address, the criminal’s capacity to respond to reasons. On our view:

By analogy, we argue that one standard form of treatment for alcoholism – which many alcoholics voluntarily undergo – involves the use of Antabuse, which makes one violently ill after the ingestion of alcohol. By counteracting addictive alcoholism, this drug can result in enhanced autonomy and reasons-responsiveness.

According to Pereboom and me, treatments and interventions that increase reasons-responsiveness in this way are ethical if carried out in a way that respects autonomy by leaving the decision up to the offender. Some ethicists worry, however, that if an offender is presented with a choice between accepting a particular intervention/drug/treatment or facing the prospect of further incarceration if they refuse, their consent is not valid since it is partly coerced. While this is a difficult issue that deserves more attention than I can give it here, my short reply would be that, although incarcerated offenders clearly face pressure to consent in such situations, that pressure does not render their consent invalid because their choice is still sufficiently voluntary (Reference Bomann-LarsenBomann-Larsen 2011; Reference Wertheimer and MillerWertheimer and Miller 2014; Reference RybergRyberg 2019). In fact, there is some evidence that the conditions under which crime-preventing interventions are agreed to by incarcerated offenders are not perceived as coercive by the offenders themselves (Reference Poythress, Petrila, McGaha and BoothroydPoythress et al. 2002; Reference RiggRigg 2002; Reference Moser, Arndt and KanzMoser et al. 2004; Reference Redlich, Hoover, Summers and SteadmanRedlich et al. 2010; Reference Douglas, Bonte, Focquart, Devolder and SterckxDouglas et al. 2013: n4).

But what about those interventions that are mandated by the court or made a requirement for release? Even here, some ethicists have argued that mandated neurointerventions and mandated incarceration can be comparable responses to crime, thereby holding that liability to incarceration implies liability to some forms of forced interventions (Reference DouglasDouglas 2014a; Reference RybergRyberg 2019). Reference DouglasThomas Douglas (2014a), for instance, questions the need to obtain informed consent for using neurointerventions (or what he calls “medical correctives”) on offenders. He writes: “Why is it that, following offending, consent is required for the imposition of medical correctives, but not for these more traditional kinds of criminal remedy [incarceration, psychological rehabilitation programs, fines, community service, probation regimes, and the freezing of financial assets]?” (Reference Douglas2014a: 105). According to Douglas:

Douglas’s argument, therefore, maintains that if the goals of rehabilitation, crime prevention, and security can justify mandatory minimal incarceration, they can also justify some nonconsensual neurointerventions.

Given that most academics debating the ethics of neurointerventions share the assumption that administering interventions to offenders without their valid consent would be unethical – disagreeing only over whether offenders can, in fact, validly consent in certain coercive circumstances – Douglas’s challenge to the “consent requirement” (Reference Douglas2014a: 104) is a significant one. If it succeeds, it would have major implications for the nonconsensual use of neurointerventions. There are, however, a number of powerful objections to this view (see Reference Shaw, Birks and DouglasShaw 2018b, Reference Shaw2019; Reference Choy, Focquaert and RaineChoy, Focquaert, and Raine 2020).

First, forced neurointerventions violate an individual’s right to bodily integrity, and the forced nature of the intervention, in and of itself, can be experienced as invasive, degrading, and humiliating (Reference ShawShaw 2019; Reference Choy, Focquaert and RaineChoy, Focquaert, and Raine 2020: 34). Elizabeth Reference ShawShaw (2019), for instance, questions whether even the supposedly benign proposal of compulsory injections is unlikely to cause serious harm. She writes:

To further complicate matters, given the wide discrepancy in human experiences, “it will be nearly impossible to outline general rules of practice based on the level of intrusiveness and harm that either incarceration or biological intervention entails” (Reference Choy, Focquaert and RaineChoy, Focquaert, and Raine 2020: 35).

Second, Reference Shaw, Birks and DouglasShaw (2018b, Reference Shaw2019) argues that the mandatory injection of prisoners poses a communicative threat to agency – that is, “Pinning someone down and forcibly injecting her with a mind-altering drug is likely to send out a more disrespectful message about the individual than incarceration would” (Reference Shaw2019: 104). She goes on to argue:

The problem is that suggesting that consent for neurointerventions is the default ethical requirement for all competent adults under normal circumstances, but not for offenders, “signals to offenders that they are somehow less deserving of respect as persons compared to nonoffenders” (Reference Choy, Focquaert and RaineChoy, Focquaert, and Raine 2020: 35). In my opinion, this is a powerful reason to object to nonconsensual interferences with both bodily and mental integrity.

For the foregoing reasons, I oppose the use of mandated or forced neurointerventions. The best way to respect the autonomy of competent adults, I contend, is to leave decisions about neurointerventions up to the individual. Mandated neurointerventions interfere with the rights of mental and bodily integrity – and violating mental and bodily integrity is generally more disrespectful and harder to justify than interfering with freedom of movement (Reference Shaw, Birks and DouglasShaw 2018b, Reference Shaw2019). While I understand that concerns about consent still remain for incarcerated offenders, it’s not clear that leaving the option up to the individual for safe and effective neurointerventions should be ruled out as morally illegitimate. This is because the offender’s choice is, I contend, still sufficiently voluntary in most circumstances. This is especially true when the neurointerventions in question are noninvasive and rather benign, like neurofeedback therapy.

But what about those neurointerventions that are more invasive, like chemical castration? And what if the only legitimate alternative to continued incarceration is for the offender to choose chemical castration? For me, a lot will depend on how safe and effective chemical castration is ultimately proven – which remains an open empirical question (Reference Rice and HarrisRice and Harris 2003). Some small-scale, controlled studies have found that chemical castration, administered using cyproterone acetate (CPA) and medroxyprogesterone acetate (MPA), is effective in reducing recidivism in sexual offenders with paraphilias (Reference Fedoroff, Wisner-Carlson, Dean and BerlinFedoroff et al. 1992; Reference Maletzky, Tolan and McFarlandMaletzky, Tolan, and McFarlan 2006). However, other studies have found no significant effect (Reference Hucker, Langevin and BainHucker, Langevin, and Bain 1988). On the other hand, “several studies have shown high efficacy rates for GnRH agonists in dramatically reducing testosterone levels and self-reported deviant sexual desires and behaviors, including in individuals who did not respond to CPA or MPA” (Reference Douglas, Bonte, Focquart, Devolder and SterckxDouglas et al. 2013: 395). But even here, more empirical studies are needed. Assuming for the moment, though, that chemical castration was proven sufficiently safe and effective, it’s not clear that under such circumstances the moral problems with it are not outweighed if it is carried out in a way that respects autonomy by leaving the decision up to the individual.

3.4 Personal Identity

Another common concern about neurointerventions has to do with personal identity. Neurointerventions that alter brain function or modify cognitive processes may raise questions about the authenticity of a person’s thoughts, emotions, or behaviors. If these interventions significantly change or manipulate core aspects of an individual’s identity, it can lead to a sense of inauthenticity or perceived loss of self. Furthermore, altering brain function through neurointerventions can potentially disrupt the continuity of a person’s identity. If the interventions cause a significant shift in personality, memories, or cognitive abilities, it may challenge the individual’s sense of continuity over time, potentially leading to a feeling of disconnection from their past self. For these reasons, “One of the most salient worries related to personal identity is the fear of creating a new person, or radically changing a person’s self up to the point where they can no longer consider themselves the same” (Reference Focquaert and DeRidderFocquaert and DeRidder 2009: 1). It’s important, then, that we examine whether neurointerventions threaten our personal identity and if the possibility of identity changes provides a sound ethical argument against these techniques.

Before answering this question directly, it’s important to note that a good deal of personal identity-related worries regarding neurointerventions rest upon a conflation of narrative and numerical identity. As Focquaert and DeRidder explain:

Focquaert and DeRidder go on to claim that: “While it is self-evident that altering an individual’s numerical identity is wrong, it is much less clear that altering one’s narrative identity is ethically problematic” (Reference Focquaert and DeRidder2009: 2). This is because, while radical narrative changes are potentially problematic, mild identity changes are not necessarily problematic. In fact, “Mild and moderate narrative identity changes are part of our daily life and may result from a variety of life-changing experiences or circumstances.”

For example, the “loss of a loved one may change someone from being optimistic to being depressed and without hope.” Of course, narrative identity changes induced by neurointerventions come about differently, “but [they] are essentially the same as those that come about just by living our daily lives” (Reference Focquaert and DeRidderFocquaert and DeRidder 2009: 2). Whether or not such changes are problematic depends on how much change in narrative identity (including changes in cognition, personality traits, emotions, and mood) is deemed “too radical” and a threat to personal identity.

Before investigating concerns about narrative identity further, let me just note that while it is self-evident that altering an individual’s numerical identity is wrong, there is also a good reason to think that personal identity through time (numerical identity) is not threatened by neurointerventions in the same way as narrative identity. Numerical identity revolves around the necessary and sufficient conditions for a person at one point in time to be the same person at another point in time. And while philosophers do not fully agree on what these conditions are, two leading approaches focus on psychological and biological continuity:

There is no reason to think, however, that neurointerventions threaten continuity in either of these senses (see Reference GlannonGlannon 2007; Reference Focquaert and DeRidderFocquaert and DeRidder 2009) since neurointerventions do not affect the continuity of one’s biological life and are “unlikely [to] affect the continuity of one’s psychological connectedness or one’s basic psychological capacities” (Reference Focquaert and DeRidderFocquaert and DeRidder 2009: 2). We can say, then, that “Philosophical and ethical worries about personal identity are typically about cases in which [neurointerventions are] successful in alleviating the patient’s symptoms, but at the same time lead to changes in one’s mental states (e.g., changes in personality traits)” (Reference Focquaert and DeRidderFocquaert and DeRidder 2009: 2). Such worries are therefore about changes in narrative identity rather than numerical identity (Reference GlannonGlannon 2007).

What, then, can current findings on neurointerventions tell us about their effects on cognition, personality traits, emotions, and mood? That is, “which aspects of our narrative identity are (likely or unlikely) altered due to stimulation for neurological and neuropsychiatric disorders? Do these alterations involve drastic changes? And if yes, are such changes common?” (Reference Focquaert and DeRidderFocquaert and DeRidder 2009: 2). To help answer these questions, consider the use of DBS to treat Parkinson’s disease. We can use this as a test case to see what personality changes, if any, are produced by DBS. Focquaert and DeRidder nicely summarize the key empirical findings as follows:

None of these personality changes seem “radical” enough to constitute a significant threat to narrative personal identity. In fact, these findings show that “mild to moderate changes in one’s narrative identity are observed in individual cases, while radical narrative changes are rare” (Reference Focquaert and DeRidderFocquaert and DeRidder 2009: 4). The same seems to be true for the use of DBS to treat neuropsychiatric disorders (see Reference Focquaert and DeRidderFocquaert and DeRidder 2009). It would therefore seem that changes in personality traits, emotions, and mood due to DBS are relatively uncommon in both Parkinson’s disease and neuropsychiatric disorders.

Similar assessments would need to be done for other neurointerventions to determine whether or not they result in the kind of radical changes that threaten narrative personal identity. For most forms of noninvasive neurointerventions, like neurofeedback and transcranial magnetic stimulation, there’s good reason to think that this will not be the case. Neurofeedback, for instance, does not radically change an individual’s personality; it simply trains their brain to work more effectively and enhance their overall mental balance and mood. Reference Dalkner, Unterrainer and WoodDalkner et al. (2017), for example, studied the short-term beneficial effects of twelve sessions of neurofeedback on avoidant personality accentuationFootnote ³ in the treatment of alcohol use disorder and found that, while neurofeedback intervention had a positive effect on avoidant personality accentuation, there were no changes in other personality accentuations or in global Big Five personality dimensions after neurofeedback training. Reference Raymond, Varney, Parkinson and GruzelierRaymond et al. (2005) similarly found that nine sessions of neurofeedback for participants with high scores for withdrawal led to improvements in mood (i.e., feeling more composed, agreeable, elevated, and confident) but were “insufficient to change personality” (Reference Raymond, Varney, Parkinson and Gruzelier2005: 287). Similarly, research on the relationship between personality traits and responses to Transcranial Magnetic Stimulation (TMS) treatments has found that although certain personality traits using the five-factor personality assessment were correlated with clinical remission (i.e., elimination) of depression symptoms, there were no changes in personality measures following a four-week course of TMS (Reference McGirr, Van den Eynde, Chachamovich and FleckMcGirr et al. 2014).

None of this is to say, of course, that neurointerventions can never threaten personal identity, since there very may well be brain implants or neurosurgeries that do or can result in radical changes in narrative personal identity. It has been widely reported, for instance, that the historical use of lobotomies to treat psychiatric disorders, while not only dangerous, often resulted in radical changes in personality – that is, many people lost their ability to feel emotions and became apathetic, unengaged, and unable to concentrate (some even became catatonic) (see Reference Caruso and SheehanCaruso and Sheehan 2017). If we want to be certain that we are using neurointerventions in an ethical manner, we therefore need to be constantly assessing the empirical effects of neurointerventions on personality traits, emotions, and mood, so as to avoid neurointerventions that bring about radical changes in narrative identity.

4.1 Neuroscience Is Already Revising the Law

To begin, it is important to recognize that developments in neuroscience have already begun to bring about changes and revisions in the law. One historical example can be found in the mid-twentieth century with the use of electroencephalography (EEG) to expand the rights of epileptics. As Francis Shen writes, “Epilepsy had ‘plagued mankind from time immemorial,’ and twenty-eight states had sterilization laws that included epileptics. As late as 1956, some states still had laws that restricted marriage of epileptics. But the ‘discovery [of EEG] … released epilepsy from the crypt of the unknown’” (Reference Shen2016: 676). Before EEG, and until the 1950s, individuals with epilepsy were legally denied the right to marry, the right to drive a car, and the right to obtain employment. Gradually, such laws targeting epileptics came under attack as lawyers and doctors banded together. These changes in the law “owed much to the discovery and application of EEG.” Early on, EEGs were particularly useful in the diagnosis and treatment of epilepsy and related seizure disorders. And in 1956, a landmark book by Roscoe Barrow and Howard Fabing was published named Epilepsy and Law: A Proposal for Legal Reform in the Light of Medical Progress. Publications such as this contributed to a series of legal reforms, “including giving epileptics the right to marry and drive cars with fewer restrictions.” The reform of laws governing epilepsy clearly demonstrates that advances in neuroscience “can lead to positive legal and social outcomes” (Reference ShenShen 2016: 679).

Another area where neuroscientific findings have, and are continuing to have, an effect on the law is in juvenile justice policy. Developmental psychologists Laurence Steinberg and Elizabeth Scott have argued, for instance, that juveniles are “less guilty by reason of adolescence” (Reference Steinberg and Scott2003: 1009). Under principles of criminal law, culpability is mitigated when the agent’s decision-making capacity is diminished, when the criminal act was coerced, or when the act was out of character. Steinberg and Scott argue that juveniles should not be held to the same standards of criminal responsibility as adults because adolescents’ decision-making capacity is diminished, they are less able to resist coercive influence, and their character is still undergoing change. The uniqueness of immaturity as a mitigating condition, they argue, suggests “commitment to a legal environment under which most youths are dealt with in a separate justice system and none are eligible for capital punishment.”

The American Psychological Association (APA) agrees with this assessment and has partnered with jurisdictions nationwide to revise juvenile sentencing policies to align with developmental science (Reference AbramsAbrams 2022). Their efforts have led to a series of amicus briefs, heavily cited in the US Supreme Court’s Roper v. Simmons decision – which said using the death penalty before age eighteen was unconstitutional – and the subsequent Graham v. Florida, Miller v. Alabama, and Jackson v. Hobbs decisions, which collectively outlawed sentencing people under eighteen to life in prison without the possibility of parole. In the Roper case, the Court admitted MRI and other neuroscientific evidence showing the absence of frontal lobe maturation in the brains of teenagers (frontal lobes are causally implicated in decision-making and the control of impulse reactions). The Court cited three relevant ways that adolescents differ from adults: lack of maturity, increased impulsivity, and limited judgment; increased vulnerability and susceptibility to external pressure and negative influences; and a personality structure that is less fixed and more open to change. This led the US Supreme Court to rule that the Eighth and Fourteenth Amendments forbid the execution of offenders who were younger than age eighteen when the crime occurred.

The APA is now working to extend these legal reforms further. Drawing on an updated body of research on developmental psychology and neuroscience, a recent APA Presidential Task Force recommended the extension of these protections beyond age 18. In August 2022, the APA Council of Representatives approved the policy resolution with overwhelming support. “The fundamental argument is that the research the Roper court relied on to exclude seventeen-year-olds from eligibility for death as a penalty also applies at ages 18, 19, and 20,” said Cecil Reynolds, the leader of the APA task force (as quoted in Reference AbramsAbrams 2022: 49). In addition to the policy brief, the task force is creating an amicus brief for the Supreme Court and a tool kit for practitioners working with 18-, 19-, and 20-year-olds facing a death penalty charge. Psychologists and neuroscientists have also summarized research on brain development and trauma to inform other policy initiatives, including advocating for states to ban solitary confinement of juveniles and to raise the minimum age for trying children in the juvenile justice system.

4.2 Free Will and Its Relevance to the Law

In addition to laws governing epilepsy and juvenile justice, some theorists have proposed that advances in neuroscience will result in even more radical revisions to the law. These theorists argue that as our neuroscientific grasp of complex behavior improves, it will undermine our commonsense notions of free will and moral responsibility. Reference Green and CohenJoshua Greene and Jonathan Cohen (2004) have argued, for instance, that the advance of neuroscience will eventually result in the widespread rejection of free will – and, with it, of retributivism. In their now classic paper, they argue:

According to Greene and Cohen, “The net effect of this influx of scientific information, will be the rejection of free will as it is ordinarily conceived, with important ramifications for the law” (Reference Green and Cohen2004: 1776). In particular, they argue that neuroscience will dramatically and beneficially change our legal system by forcing it to take cognizance of developments in our understanding of human capacities. This process will move it and us away from our retributive urges and toward a more compassionate consequentialist form of punishment in the future. They go on to propose that consequentialist reforms are in order, and they predict such reforms will take place.

But before examining the various neuroscientific and philosophical arguments against free will, it’s important that we first acknowledge and fully appreciate the relevance of free will to the criminal law. First and foremost, the criminal law is founded on the idea that persons can be held morally responsible for their actions because they have freely chosen them. The US Supreme Court, for instance, has asserted:

Indeed, the US courts have observed that “[t]he whole presupposition of the criminal law is that most people, most of the time, have free will within broad limits”Footnote ⁴ and that “the law has been guided by a robust common sense which assumes the freedom of the will as a working hypothesis in the solution of its problems.”Footnote ⁵

Free will is also relevant to criminal law because of the important role it plays, historically and currently, in one of the most prominent justifications of legal punishment: retributivism. The retributive justification of legal punishment maintains that, absent any excusing conditions, wrongdoers are morally responsible for their actions and deserve to be punished in proportion to their wrongdoing. Unlike theories of punishment that aim at deterrence, rehabilitation, or incapacitation, retributivism grounds punishment in the blameworthiness and desert of offenders. It holds that punishing wrongdoers is intrinsically good. For the retributivist, wrongdoers deserve a punitive response proportional to their wrongdoing, even if their punishment serves no further purpose. This backward-looking focus on desert is a central feature of most pure retributive accounts of punishment. And it is important to emphasize that the desert invoked in retributivism (in the classical or strict sense) is basic in the sense that it is not in turn grounded in forward-looking reasons such as securing the safety of society or the moral improvement of criminals. Thus, for the retributivist, the claim that persons are morally responsible for their actions in the basic desert sense is crucial to the state’s justification for giving them their just deserts in the form of punishment for violations of the state’s laws.

Given the importance, then, that assumptions about free will and basic desert moral responsibility play in the criminal law, if it turns out that agents lack free will, then major aspects of our criminal law and justice systems are unjustified. It’s imperative, then, that we examine the neuroscientific and philosophical arguments against free will since one reason to think that no one ever deserves legal punishment, as retributivism presupposes, is that we lack the control in action, that is, the free will, required for moral responsibility in the basic desert sense.

4.3 Neuroscience and Free Will

In the literature, two prominent routes to free will skepticism are identifiable. The first, which is more prominent among scientific skeptics, maintains that recent findings in neuroscience reveal that unconscious brain activity causally initiates action prior to the conscious awareness of the intention to act, and that this indicates conscious will is an illusion. The pioneering work in this area was done by Benjamin Libet and his colleagues. In their groundbreaking study on the neuroscience of movement, Reference Libet, Gleason, Wright and PearlLibet et al. (1983) investigated the timing of brain processes and compared them to the timing of conscious will in relation to self-initiated voluntary acts, and found that the conscious intention to move (which they labeled W) came 200 milliseconds before the motor act, but 350–400 milliseconds after the readiness potential (RP) – a ramplike buildup of electrical activity that occurs in the brain and precedes actual movement. Libet and others have interpreted this as showing that the conscious intention or decision to move cannot be the cause of action because it comes too late in the neuropsychological sequence (Reference LibetLibet 1985, Reference Libet1999). According to Libet, since we become aware of an intention to act only after the onset of preparatory brain activity, the conscious intention cannot be the true cause of the action.

Libet’s findings, in conjunction with additional findings by John Dylan Haynes (Reference Soon, Brass, Heinze and HaynesSoon et al. 2008) and others, have led some theorists to conclude that conscious will is an illusion and plays no important causal role in how we act (see Reference WegnerWegner 2002). Haynes and his colleagues, for example, were able to build on Libet’s work by using functional magnetic resonance imaging (fMRI) to predict with 60 percent accuracy whether subjects would press a button with either their right or left hand up to ten seconds before the subject became aware of having made that choice (Reference Soon, Brass, Heinze and HaynesSoon et al. 2008). For some, the findings of Libet and Haynes are enough to threaten our conception of ourselves as free and responsible agents since they appear to undermine the causal efficacy of the types of willing required for free will.

There are, however, at least three reasons for thinking that these neuroscientific arguments for free will skepticism are unsuccessful.Footnote ⁶ First, there is no direct way to tell which conscious phenomena, if any, correspond to which neural events. In particular, in the Libet studies, it is difficult to determine what the RP corresponds to – for example, is it an intention formation or decision, or is it merely an urge of some sort? Reference MeleAl Mele (2009) has argued that the RP that precedes action by a half-second or more need not be construed as the cause of the action. Instead, it may simply mark the beginning of forming an intention to act. According to Mele, “it is much more likely that what emerges around –500 ms is a potential cause of a proximal intention or decision than a proximal intention or decision itself” (Reference Mele2009: 51). On this interpretation, the RP is more accurately characterized as an “urge” to act or a preparation to act. I agree with Mele that this leaves open the possibility that conscious intentions can still be causes – that is, if the RP does not correspond to the formation of an intention or decision, but rather an urge, then it remains open that the intention formation or decision is a conscious event.

Second, almost everyone on the contemporary scene who believes we have free will, whether compatibilist or libertarian, also maintains that freely willed actions are caused by virtue of a chain of events that stretch backward in time indefinitely. At some point in time, these events will be such that the agent is not conscious of them. Thus, all free actions are caused, at some point in time, by unconscious events. However, as Reference Nahmias and KaneEddy Nahmias (2011) correctly points out, the concern for free will raised by Libet’s work is that all of the relevant causing of action is (typically) nonconscious, and consciousness is not causally efficacious in producing action. Given determinist compatibilism, however, it is not possible to establish this conclusion by showing that nonconscious events that precede conscious choice causally determine action since such compatibilists hold that every case of action will feature such events and that this is compatible with free will. And, given most incompatibilist libertarianisms, it is also impossible to establish this conclusion by showing that there are nonconscious events that render actions more probable than not by a factor of 10 percent chance (Reference Soon, Brass, Heinze and HaynesSoon et al. 2008) since almost all such libertarians hold that free will is compatible with such indeterminist causation by unconscious events at some point in the causal chain.

Furthermore, Neil Levy raises a related objection when he criticizes Libet’s impossible demand (Reference Levy2005) that only consciously initiated actions could be free. Levy correctly argues that this presupposition places a condition upon freedom of action which is, in principle, impossible to fulfill for reasons that are entirely conceptual and have nothing to do, per se, with Libet’s empirical findings. As Levy notes, “Exercising this kind of control would require that we control our control system, which would simply cause the same problem to arise at a higher-level or initiate an infinite regress of controllings” (Reference Levy2005: 67). If the unconscious initiation of actions is incompatible with control over them, then free will is impossible on conceptual grounds. Thus, Libet’s experiments do not constitute a separate, empirical challenge to our freedom.

Finally, several critics have correctly noted the unusual nature of the Libet-style experimental situation – that is, one in which a conscious intention to flex at some time in the near future is already in place, and what is tested for is the specific implementation of this general decision. Reference Nahmias and KaneNahmias (2011), for example, convincingly points out that it is often the case – when, for instance, we drive or play sports or cook meals – that we form a conscious intention to perform an action of a general sort, and subsequent specific implementations are not preceded by more specific conscious intentions. But, in such cases, the general conscious intention is very plausibly playing a key causal role. In Libet-style situations, when the instructions are given, subjects form conscious intentions to flex at some time or other, and, if it turns out that the specific implementations of these general intentions are not in fact preceded by specific conscious intentions, this would be just like the kinds of driving and cooking cases Nahmias cites. It seems that these objections cast serious doubts on the potential for neuroscientific studies to undermine the claim that we have the sort of free will at issue.

For the foregoing reasons, I reject this kind of neuroscientific argument against free will. Of course, doing so does not mean that there are not other, better arguments against free will. In fact, I will now present what I take to be the strongest and most compelling case for free will skepticism – one that does not appeal to epiphenomenalism or the timing of conscious processes.

4.4 Hard-Incompatibilism

In the past, the standard argument for free will skepticism was based on the notion of determinism – the thesis that facts about the remote past in conjunction with the laws of nature entail that there is only one unique future. Hard determinists argued that determinism is true and incompatible with free will and basic desert moral responsibility, either because it precludes the ability to do otherwise (leeway incompatibilism) or because it is inconsistent with one’s being the “ultimate source” of action (source incompatibilism). For hard determinists, libertarian free will is an impossibility because human actions are part of a fully deterministic world and compatibilism is operating in bad faith. The hard determinist invokes the naturalistic consideration that everything that happens, including all of our actions, is made inevitable by the remote past physical events together with the laws of nature. It maintains that human beings are situated in a natural world of law-governed causes and effects, and as a result our character and actions are conditioned by causes that we do not control, including our genetic make-up, our upbringing, and the physical environment. Because of these general features of the universe, the hard determinist maintains that we cannot have the sort of control in action required for basic desert attributions.

While hard determinism had its classic statement in the time when Newtonian physics reigned, it has very few defenders today – largely because the development of quantum mechanics diminished confidence in determinism, for the reason that it has indeterministic interpretations. This is not to say that determinism has been refuted or falsified by modern physics, because it has not. In fact, a number of leading interpretations of quantum mechanics are consistent with determinism (see Reference BohmBohm 1952; Reference VaidmanVaidman 2014; Reference LewisLewis 2016). It is also important to keep in mind that even if we allow some indeterminacy to exist at the microlevel of the universe, the level studied by quantum mechanics, there may still remain determinism-where-it-matters – that is, at the ordinary level of choices and actions, and even the electrochemical activity in our brains. Nevertheless, most contemporary free will skeptics now offer arguments that are agnostic about determinism (see Reference PereboomPereboom 2001, Reference Pereboom2014; Reference LevyLevy 2011; Reference StrawsonStrawson 1986; Reference CarusoCaruso 2012, Reference Caruso2021a).

My own variety of free will skepticism maintains that whatever the fundamental nature of reality – whether it be deterministic or indeterministic – we lack free will (Reference PereboomPereboom 2001, Reference Pereboom2014; Reference CarusoCaruso 2012, Reference Caruso2021a). This is because, while determinism is incompatible with free will, so too is any sort of indeterminism that has a good chance of being true. A more accurate name for this position would therefore be hard incompatibilism (Reference PereboomPereboom 2001, Reference Pereboom2014), to differentiate it from hard determinism. Hard incompatibilism amounts to a rejection of both compatibilism and libertarianism. It maintains that the sort of free will required for basic desert moral responsibility is incompatible with causal determination by factors beyond the agent’s control and also with the kind of indeterminacy in action required by the most plausible versions of libertarianism.

I contend that the most direct and convincing route to this conclusion runs as follows. Against the view that free will is compatible with the causal determination of our actions by natural factors beyond our control, I argue that there is no relevant difference between this prospect and our actions being causally determined by manipulators. Against libertarian views that appeal only to indeterminate events and states of the agent to preserve free will, I advance the disappearing agent objection, according to which agents are left unable to settle whether a decision occurs and hence cannot have the control required for basic desert moral responsibility. In response, some libertarians appeal to a more robust notion of causation according to which free will requires an agent, qua substance, to be irreducibly causally involved. While such accounts could, in theory, supply the control needed for basic desert moral responsibility, I argue that they cannot be reconciled with our best philosophical and scientific accounts of the world and face additional problems accounting for mental causation. Since this exhausts the options for views on which we have the sort of free will at issue, I conclude that free will skepticism is the only remaining position.

While these arguments have been defended at great length elsewhere (see Reference PereboomPereboom 2001, Reference Pereboom2014, Reference Pereboom2022; Reference CarusoCaruso 2012, Reference Caruso2021a; Reference Pereboom, Caruso, Caruso and FlanaganPereboom and Caruso 2018; Reference PereboomCaruso and Pereboom 2022), I will briefly outline them in the following section. I will also examine which aspects admit of empirical considerations and which do not, in hopes of better understanding what neuroscience can and cannot answer and where philosophy must enter the picture.

4.5 The Epistemic Argument

If what I’ve argued in the previous sections is correct, we have sufficient reason for adopting the skeptical conclusion that who we are and what we do is ultimately the result of factors beyond our control (whether those factors be deterministic or indeterministic), and because of this, we are never morally responsible for our actions in the basic desert sense. Adopting this conclusion entails that retributivism is unjustified, and major revisions to the law are required. Furthermore, this argument in no way appeals to controversial claims about the timing of conscious will or presupposes the truth of determinism – hence, it differs significantly from the kind of neuroscientific argument against free will discussed earlier. It also differs from the kind of argument provided by Reference Green and CohenGreene and Cohen (2004), since their argument relies heavily on intuitions about neurobiological determinism and the assumption that the criminal law is committed to an untenable libertarian account of agency. The hard incompatibilist argument for revision relies on neither of these things. Instead, it contends that whether or not the law assumes a libertarian or compatibilist notion of free will, it nonetheless is in need of revision – since there are powerful and insurmountable arguments against both kinds of free will, and neither is able to preserve the control in action required for basic desert moral responsibility and retributive punishment.

But what if one is still not convinced by the arguments for hard incompatibilism? Well, I maintain that even in the face of uncertainty about the existence of free will, it remains unclear whether retributive punishment is justified. This is because the burden of proof lies on those who want to inflict harm on others to provide good justification for such harm. This means that retributivists who want to justify legal punishment on the assumption that agents are free and morally responsible (and hence justly deserve to suffer for the wrongs they have done) must justify that assumption. And they must justify that assumption in a way that meets a high epistemic standard of proof since the harms caused in the case of legal punishment are often quite severe. It is not enough to simply point to the mere possibility that agents possess libertarian or compatibilist free will. Nor is it enough to say that the skeptical arguments against free will and basic desert moral responsibility fail to be conclusive. Rather, a positive and convincing case must be made that agents are in fact morally responsible in the basic desert sense, since it is the backward-looking desert of agents that retributivists take to justify the harm caused by legal punishment.

I call this second argument against retributive legal punishment the Epistemic Argument (Reference CarusoCaruso 2020a, Reference Caruso2021a), and it can be summarized as follows:Footnote ⁸

1. 1. Legal punishment intentionally inflicts harms on individuals and the justification for such harms must meet a high epistemic standard. If it is significantly probable that one’s justification for harming another is unsound, then, prima facie, that behavior is seriously wrong.

2. 2. The retributive justification for legal punishment assumes that agents are morally responsible in the basic desert sense and hence justly deserve to suffer for the wrongs they have done in a backward-looking, non-consequentialist sense (appropriately qualified and under the constraint of proportionality).

3. 3. If the justification for the assumption that agents are morally responsible in the basic desert sense, and hence justly deserve to suffer for the wrongs they have done, does not meet the high epistemic standard specified in (1), then retributive legal punishment is prima facie seriously wrong.

4. 4. The justification for the claim that agents are morally responsible in the basic desert sense, provided by both libertarians and compatibilists, faces powerful and unresolved objections. As a result, it falls far short of the high epistemic bar needed to justify such harms.

5. 5. Hence, retributive legal punishment is unjustified, and the harms it causes are prima facie seriously wrong.

Note that the Epistemic Argument requires only a weaker notion of skepticism than the one defended in the previous section, namely one that holds that the justification for believing that agents are morally responsible in the basic desert sense is too weak to justify the intentional suffering caused by retributive legal punishment. Unlike the arguments for hard incompatibilism, which aim to establish that no one is ever morally responsible for their actions in the basic desert sense, the Epistemic Argument does not require the refutation of libertarian and compatibilist accounts of free will. Instead, it simply needs to raise sufficient doubt that they succeed. And such doubt, I claim, is easy to raise. As a result, we should conclude that retributive legal punishment is unjustified and the harms it causes are prima facie seriously wrong (see Reference CarusoCaruso 2021a for more details).

4.6 Revision

If we come to doubt or deny the existence of free will, or reject retributivism for other reasons, where does that leave us with regard to criminal justice? Well, traditionally, in addition to pure retributivism, there have been a number of other common justifications for legal punishment. Greene and Cohen, for instance, write: “There are perfectly good, forward-looking justifications for punishing criminals that do not depend on metaphysical fictions” (Reference Green and Cohen2004: 1783). Consequentialist deterrence theories have probably been the most discussed of these forward-looking alternatives. These theories maintain that we should only punish wrongdoers when it is rational to expect that it would maximize utility or consequentialist value relative to all the other options. These future benefits primarily include deterrence and increased safety. And the capacity to deter can further be divided into two different types: general deterrence and specific (or special) deterrence. General deterrence can be defined as the deterrence achieved from the threat of legal punishment on the public at large. Specific, or special, deterrence is deterrence aimed at previous offenders in order to reduce the likelihood of their re-offending.

While consequentialist theories are completely compatible with free will skepticism, some critics have argued that they suffer from their own independent moral difficulties (see Reference BooninBoonin 2008; Reference ZimmermanZimmerman 2011; Reference PereboomPereboom 2014; Reference CarusoCaruso 2021a). First, critics object that deterrence theories have the potential to justify punishments that are intuitively too severe. This is because, in certain cases, harsh punishments would be more effective deterrents than milder forms. Second, such accounts could potentially justify punishing the innocent. Pereboom provides the following example: “If after a series of horrible crimes the actual perpetrator is not caught, potential criminals might come to believe that they can get away with serious wrongdoing. Under such circumstances it might maximize utility to frame and punish an innocent person” (Reference Pereboom2014: 164). Lastly, there is the “use” objection, which is a problem for consequentialism more generally – that is, consequentialism “sometimes requires people to be harmed severely, without their consent, in order to benefit others, and this is often intuitively wrong” (Reference PereboomPereboom 2014: 165). While some theorists think consequentialist deterrence theories can overcome these ethical concerns (see Reference Green and CohenGreene and Cohen 2004; Reference Bennett and AltmanBennett 2023), I prefer to avoid them by adopting an altogether different approach.

I maintain that there is an ethically defensible and practically workable alternative for dealing with dangerous crime that is not undercut by either free will skepticism or by other moral considerations. It is the public health-quarantine model developed and defended by me and Derk Pereboom (Reference PereboomPereboom 2001, Reference Pereboom2014, Reference Pereboom2021; Reference CarusoCaruso 2016, Reference Caruso2017, Reference Caruso2021a, Reference Caruso2021b; Reference Pereboom, Caruso, Caruso and FlanaganPereboom and Caruso 2018; Reference Caruso, Pereboom, Focquaert, Waller and ShawCaruso and Pereboom 2020). The core idea of the model is that the right to harm in self-defense and defense of others justifies incapacitating the criminally dangerous with the minimum harm required for adequate protection. The theory is based on an analogy with quarantine and draws on a comparison between the treatment of dangerous criminals and the treatment of carriers of dangerous diseases. In its simplest form, it can be stated as follows: (1) Free will and basic desert skepticism maintain that criminals are not morally responsible for their actions in the basic desert sense; (2) plainly, many carriers of dangerous diseases are not responsible in this or in any other sense for having contracted these diseases; (3) yet, we generally agree that it is sometimes permissible to quarantine them, and the justification for doing so is the right to self-protection and the prevention of harm to others; (4) for similar reasons, even if a dangerous criminal is not morally responsible for his crimes in the basic desert sense (perhaps because no one is ever in this way morally responsible) it could be as legitimate to preventatively detain him as to quarantine the non-responsible carrier of a dangerous disease.

The first thing to note about the theory is that although one might justify quarantine (in the case of disease) and incapacitation (in the case of dangerous criminals) on purely utilitarian or consequentialist grounds, Pereboom and I resist this strategy. Instead, we maintain that incapacitation of the seriously dangerous is justified on the ground of the right to harm in self-defense and defense of others. That we have this right has broad appeal, much broader than utilitarianism or consequentialism more generally has. In addition, this makes the view more resilient to a number of objections and provides a more robust proposal for justifying criminal sanctions than other non-retributive options (Reference CarusoCaruso 2021a, Reference Caruso2021b). One advantage it has, say, over consequentialist deterrence theories, is that it has more restrictions placed on it with regard to using people merely as a means. For instance, as it is illegitimate to treat carriers of a disease more harmfully than is necessary to neutralize the danger they pose, treating those with violent criminal tendencies more harshly than is required to protect society will be illegitimate as well. In fact, the model requires that we adopt the principle of least infringement, which holds that the least restrictive measures should be taken to protect public health and safety. This ensures that criminal sanctions will be proportionate to the danger posed by an individual, and any sanctions that exceed this upper bound will be unjustified.

Second, the quarantine model places several constraints on the treatment of criminals. First, as less dangerous diseases justify only preventative measures less restrictive than quarantine, so less dangerous criminal tendencies justify only more moderate restraints. We do not, for instance, quarantine people for the common cold, even though it has the potential to cause you some harm. Rather, we restrict the use of quarantine to a narrowly prescribed set of cases. Analogously, on this model, the use of incapacitation should be limited to only those cases where offenders are a serious threat to public safety and no less restrictive measures were available. In fact, for certain minor crimes, perhaps only some degree of monitoring could be defended. Second, the incapacitation account that results from this analogy demands a degree of concern for the rehabilitation and well-being of the criminal that would alter much of current practice. Just as fairness recommends that we seek to cure the diseased we quarantine, so fairness would counsel that we attempt to rehabilitate the criminals we detain. Rehabilitation and reintegration would therefore replace punishment as the focus of the criminal justice system. Lastly, if a criminal cannot be rehabilitated and our safety requires his indefinite confinement, this account provides no justification for making his life more miserable than would be required to guard against the danger he poses.

In addition to these restrictions on harsh and unnecessary treatment, the model also advocates for a broader approach to criminal behavior that moves beyond the narrow focus on sanctions. Most importantly, it situates the quarantine analogy within the broader justificatory framework of public health ethics (Reference CarusoCaruso 2016, Reference Caruso2021a). Public health ethics not only justifies quarantining carriers of dangerous diseases on the grounds that it is necessary to protect public health, but it also requires that we take active steps to prevent such outbreaks from occurring in the first place. The analogous claim holds for incapacitation. Taking a public health approach to criminal behavior allows us to justify the incapacitation of dangerous criminals when needed, but it also makes prevention a primary function of the criminal justice system. So instead of focusing on punishment, the public health-quarantine model shifts the focus to identifying and addressing the systemic causes of crime, such as poverty, low socioeconomic status, systematic disadvantage, mental illness, homelessness, educational inequity, exposure to abuse and violence, poor environmental health, and addiction.

In fact, the public health framework sees social justice as a foundational cornerstone to public health and safety (Reference CarusoCaruso 2021b). In public health ethics, a failure on the part of public health institutions to ensure the social conditions necessary to achieve a sufficient level of health is considered a grave injustice. An important task of public health ethics, then, is to identify which inequalities in health are the most egregious and thus which should be given the highest priority in public health policy and practice. The public health approach to criminal behavior likewise maintains that a core moral function of the criminal justice system is to identify and remedy social and economic inequalities responsible for crime. Just as public health is negatively affected by poverty, racism, and systemic inequality, so too is public safety. This broader approach to criminal justice therefore places issues of social justice at the forefront. It sees racism, sexism, poverty, and systemic disadvantage as serious threats to public safety and prioritizes the reduction of such inequalities.

Summarizing the public health-quarantine model, then, the core idea is that the right to harm in self-defense and defense of others justifies incapacitating the criminally dangerous with the minimum harm required for adequate protection. The resulting account would not justify the sort of criminal punishment whose legitimacy is most dubious, such as death or confinement in the common kinds of prisons in the United States. The model also specifies attention to the well-being of criminals, which would change much of the current policy. Furthermore, the public health component of the theory prioritizes prevention and social justice and aims to identify and take action on the social determinants of health and criminal behavior. This combined approach to dealing with criminal behavior, I maintain, is sufficient for dealing with dangerous criminals, leads to a more humane and effective social policy, and is actually preferable to the harsh and often excessive forms of punishment that typically come with retributivism.