Skip to main content
Log in

Risk-Based Sentencing and Predictive Accuracy

  • Published:
Ethical Theory and Moral Practice Aims and scope Submit manuscript

Abstract

The use of risk assessment tools has come to play an increasingly important role in sentencing decisions in many jurisdictions. A key issue in the theoretical discussion of risk assessment concerns the predictive accuracy of such tools. For instance, it has been underlined that most risk assessment instruments have poor to moderate accuracy in most applications. However, the relation between, on the one hand, judgements of the predictive accuracy of a risk assessment tool and, on the other, conclusions concerning the justified use of such an instrument in sentencing practice, is often very unclear. The purpose of this paper is to examine this relation. More precisely, it is argued that the relation between predictive accuracy and the question as to whether a new risk assessment tool should be introduced in sentencing practice is highly complicated. For instance, there may be cases in which a new risk assessment tool is more accurate than those currently in use, but should nevertheless not be introduced; and conversely, where a new tool is less accurate, but where its introduction instead of current tools would be morally preferable.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Similar content being viewed by others

Notes

  1. In the following, predictive accuracy is used to denote the rate in which a predictive tool succeeds in making true predictions of re-offending. Moreover, for reasons of simplicity, the ensuing discussion will only concern predictions of whether or not an offender will recidivate, leaving out considerations of probabilities. For more general methodological considerations of predictive accuracy, see for instance, Fazel et al. (2012); Fazel (2019).

  2. In recent years, the concept of “algorithmic fairness” has attracted increasing attention (see e.g. Chiao 2019; Kehl et al. 2017). The discussion concerns diverging conceptions of fairness and the extent to which these can be reconciled (see Chouldechova 2017; Kleinberg et al. 2017). For instance, one account of fairness may refer to the accuracy of algorithmic predictions, while another may concern the extent to which an algorithm leads to generalizations about particular groups. In a criminal justice context, this discussion has arisen from the fact that some risk assessment tools – such as, the COMPAS algorithm which is being used for risk-assessment in several US jurisdictions – have been held to be racially biased (see e.g. Carlson 2017; Freeman 2016). However, the arguments presented in this paper do not concern unfairness in relation to any kind of differential or discriminatory treatment. Rather, they are based exclusively on penal theoretical considerations.

  3. There may of course be various types of consequence that a relevant from a penal theoretical point of view (see e.g. Ryberg 2016). However, the arguments in the following do not hinge on a particular view on which consequences should be seen as morally relevant except for the fact that crime prevention is regarded as morally desirable. For a recent discussion of the role of consequences in retributivist penal theory, see e.g. Husak (2019) or Ryberg (2019). As argued in these two chapters, any plausible positive retributivist view would have to accept that consequences should at least be taken into account in the determination of the type of punishment that should be imposed on offenders. Even if proportionality should be observed – that is, even if one holds that the severity of a punishment should reflect the seriousness of the crime – it would still be morally highly dubious not to impose the type of punishment which has the best consequences.

  4. This question my strike some as having affinity with the traditional discussion of the so-called “Blackstone’s ratio” according to which it is better that ten guilty persons escape than that one innocent person is punished. However, it should be noted that, in the present context, we are only considering the punishment of offenders and, more importantly, the arguments presented here do not presuppose a particular principle for how one should morally compare the prevention of crimes of varying degrees of severity and the punishments of offenders with varying degree of severity. For a comprehensive discussion of Blackstone’s ratio, see Laudan 2006)

  5. For an overview, see de Keijser et al. (2019).

  6. For instance, in commenting on offenders who re-offend in Pennsylvania, Rhys Hester notes that: “The majority of offenders commit low level property, drug and public order offences. Thus, most ‘high-risk’ offenders are also low-stake offenders who do not pose a serious threat to public safety” (Hester 2019, p. 218).

  7. For instance, Richard G. Singer has underlined that it is a misconception to think of the desert model as a derivative of a “throw away the key” approach to punishment. He has suggested that, in contrast to what is practiced in many jurisdictions, confinement should be reserved only for the most serious crimes and, even then, the duration should be relatively short (Singer 1979, p. 44). Along the same lines, another retributivist, Jeffrie Murphy, holds that if desert theory were to be followed consistently one would punish less and in more decent ways than is actually done (Murphy 1979, p. 230). And many more recent retributivists hold corresponding views. For instance, Andreas von Hirsch has suggested models for scaling down punishment levels and has even contended that terms of imprisonment should seldom exceed five years (see von Hirsch 1993, 2017; von Hirsch and Ashworth 2005)

  8. For instance, Michael Tonry summarizes his review of studies on crime prevention in the following way: “In 2017, it is not controversial to assert that the crime prevention effects of mass incarceration have been much less than many people supposed or hoped, that there is little or no reason to believe that harsher punishments have greater deterrent effects than milder punishments, that incapacitating people by locking them up for lengthy periods is an ineffective crime prevention strategy, or that the experience of imprisonment makes many offenders more, not less, likely to commit crimes later in their lives” and, furthermore, that: “The implications of the literatures on deterrence and incapacitation are straightforward: few convicted offenders should be sent to prison and for shorter times” (Tonry 2016, p. 453 and 459)

  9. For various discussions of the use of predictive tools within a retributivist framework, see Keijser et al. 2019.

  10. It should be noted that this would be so, even if the longer prison term succeeded in preventing the offender from committing the crime he would have committed had he received only the more lenient punishment. From a retributivist point of view, over-punishment cannot be justified in terms of crime prevention.

  11. This could be the case if the crime that is prevented by keeping the offender in prison for a longer time would not have been very serious. However, it could also be the case if the extended prison term only implies that the offender re-offends at a later point in time than would have been the case had his prison term not been prolonged.

  12. That risk assessment might imply that some offenders will be punished more severely or leniently is not entirely unrealistic. For instance, this could be the case if it is mistakenly believed that the future crimes of positives can be prevented by giving them a longer prison term. Moreover, with regard to negatives, part of the impetus behind the introduction of risk assessment in some jurisdictions has been the wish to counter mass incarceration by directing some offenders – those who are regarded as least likely to re-offend – out of prison.

  13. From a limiting retributivist point of view, this would be the case even if SuperPred were to prevent more crimes. From the retributivist it would – at least under usual conditions – not be acceptable to violate justice out of benefit in terms of crime prevention. From a consequentialist point of view, the introduction of SuperPred would be unacceptable if the harm caused by the longer prison terms on positives outweighed the benefit from preventing certain crimes; and if the harm prevented by punishing negatives leniently is outweighed by an increase in the harm caused by an increase in the number of crimes committed.

  14. The consequentialist calculation should of course also take into account the harm that will be prevented, by punishing some offenders more severely. However, we might arguendo assume that those crimes will not be very serious or, even, that the more severe punishment will not lead to an increase in crime prevention.

  15. For the negative retributivist, as noted, there would be no lower desert limit. However, since the punishment that should be imposed on a particular offender depends upon consequentialist considerations below the upper desert level, it is obviously also the case that an offender could be under-punished from a negative retributivist point of view.

  16. The discussion in this article has focused narrowly on the use of risk assessment tools in relation to sentencing. Clearly, it is worth considering whether any of the arguments that have been presented below could be extrapolated to the use of risk assessments in other contexts. However, for obvious reasons, such a discussion would reach far beyond the scope of this article.

References

  • Carlson AM (2017) “The need for transparency in the age of predictive sentencing algorithms”, Iowa Law Review 303–330

  • Chiao V (2019) Fairness, accountability and transparency: notes on algorithmic decision-making in criminal justice. International Journal of Law in Context 15:126–139

    Article  Google Scholar 

  • Chouldechova A (2017) Fair prediction with disparate impact: a study of Bias in recidivism prediction. Big Data 5:153–163

    Article  Google Scholar 

  • Douglas T, Pugh J, Singh I, Savulescu J, Fazel S (2017) Risk assessment tools in criminal justice and forensic psychiatry: the need for better data. European Psychiatry 42:134–137

    Article  Google Scholar 

  • Fazel S et al (2012) Use of risk assessment instrument to predict violence and antisocial behaviour in 73 samples involving 24827 people: systematic review and meta-analysis. Br Med J 345:1–12

    Article  Google Scholar 

  • Fazel S (2019) The Scientific Validity of Current Approaches to Violence and Criminal Risk Assessment”, in J. de Keijser et al. (eds.) (2019), Predictive sentencing: normative and empirical perspectives, Oxford: Hart Publishing

  • Frase R (2013) Just sentencing: principles and procedures for a workable system. Oxford University Press, Oxford

    Google Scholar 

  • Freeman K (2016) Algorithmic injustice: how the Wisconsin supreme court failed to protect due process rights in the state v. Loomis. North Carolina Journal of Law and Technology 18:75–106

    Google Scholar 

  • Ginneken E (2019), “The Use of Risk Assessment in Sentencing”, in J. de Keijser et al. (eds.) (2019), Predictive sentencing: normative and empirical perspectives, Oxford: Hart Publishing

  • Hannah-Moffat K, Montford KS (2019) Unpacking Sentencing Algorithms: Risk, Racial Accountability, and Data Harms, in J. de Keijser et al. (eds.) (2019), Predictive sentencing: normative and empirical perspectives, Oxford: Hart Publishing

  • Hester R (2019) “Risk Assessment at Sentencing: The Pennsylvania Experience”, in J. de Keijser et al. (eds.) (2019), Predictive sentencing: normative and empirical perspectives, Oxford: Hart Publishing

  • Husak D (2019) “Why Legal Philosophers (Including Retributivists) Should Be Less Resistant to Risk-Based Sentencing”, in J. de Keijser et al. (eds.) (2019), Predictive sentencing: normative and empirical perspectives, Oxford: Hart Publishing

  • Keijser J et al (eds) (2019) Predictive sentencing: normative and empirical perspectives. Hart Publishing, Oxford

    Google Scholar 

  • Kleinberg J et al. (2017). “Inherent trade-offs in the fair determination of risk scores”, Proceedings of Innovation in Theoretical Computer Science: 1–23

  • Laudan L (2006) Truth, error, and criminal law. Cambridge University Press, Cambridge

    Book  Google Scholar 

  • Morris N, Tonry M (1990) Between prison and probation. Oxford University Press, New York

    Google Scholar 

  • Murphy JG (1979) Retribution, justice, and therapy. Reidel, Dordrecht

    Book  Google Scholar 

  • Ryberg J (2016) “Punishment and Political Philosophy”. In Oxford Research Encyclopedia of Politics (https://oxfordre.com/politics/view/10.1093/acrefore/9780190228637.001.0001/acrefore-9780190228637-e-200?rskey=CW1R7s&result=1)

  • Ryberg J (2019) “Risk and Retribution. On the Possibility of Reconciling Considerations of Dangerousness and Desert”, in J. de Keijser et al. (eds.) (2019), Predictive sentencing: normative and empirical perspectives, Oxford: Hart Publishing

  • Singer RG (1979) Just Deserts. Balling Publishing Company, New York

    Google Scholar 

  • Singh JP et al (2014) International perspectives on the practical application of violence risk assessment: a global survey of 44 countries. Int J Forensic Ment Health 13:193–206

    Article  Google Scholar 

  • Slobogin C (2018) Principles of risk assessment: sentencing and policing. Ohio State Journal of Criminal Law 15:583–596

    Google Scholar 

  • Slobogin C (2019) “A Defence of Modern Risk-Based Sentencing”, in J. de Keijser et al. (eds.) (2019), Predictive sentencing: normative and empirical perspectives, Oxford: Hart Publishing

  • Tonry M (1995) Proportionality, parsimony, and interchangeability of punishments. In: Duff R et al (eds) Penal theory and penal practice. Manchester University Press, Manchester

    Google Scholar 

  • Tonry M (2016) Making American sentencing just, humane, and effective. Crime Justice 46(1):441–504

    Article  Google Scholar 

  • Tonry M (2019) “Sentencing and Prediction: Old Wine in Old Bottles”, in J. de Keijser et al. (eds.) (2019), Predictive sentencing: normative and empirical perspectives, Oxford: Hart Publishing

  • von Hirsch A (1993) Censure and sanctions. Clarendon Press, Oxford

    Google Scholar 

  • von Hirsch A, Ashworth A (2005) Proportionate sentencing. Oxford University Press, Oxford

    Book  Google Scholar 

  • Yang M et al (2010) The efficacy of violence prediction: a meta-analytic comparison of nine risk assessment tools. Psychol Bull 136:740–767

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Jesper Ryberg.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Ryberg, J. Risk-Based Sentencing and Predictive Accuracy. Ethic Theory Moral Prac 23, 251–263 (2020). https://doi.org/10.1007/s10677-020-10066-3

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10677-020-10066-3

Keywords

Navigation