Abstract
The idea of using artificial intelligence as a support system in the sentencing process has attracted increasing attention. For instance, it has been suggested that machine learning algorithms may help in curbing problems concerning inter-judge sentencing disparity. The purpose of the present article is to examine the merits of this possibility. It is argued that, insofar as the unfairness of sentencing disparity is held to reflect a retributivist view of proportionality, it is not necessarily the case that increasing inter-judge uniformity in sentencing is desirable. More generally, it is shown that the idea of introducing machine learning algorithms, that produce sentencing predictions on the ground of a dataset that is built of previous sentencing decisions, faces serious problems if there exists a discrepancy between actual sentencing practice and the sentences that are ideally desirable.
Similar content being viewed by others
Notes
For a recent study of sentencing disparity within the same jurisdiction in Australia, see Farmer et al. (2017).
For an overview of some of the early systems, see Schild (1998).
For instance, von Hirsch holds that “A sentencing system should seek to be just—or at least, to be as little unjust at possible. Claims about fairness … underline the requirements of proportionality” (von Hirsch 1993, p. 103).
Some theorists in the retributivist tradition would hold that desert theory does not prescribe precisely how different crimes should be punished. Rather, considerations of desert only place upper constraints on how severely an offender may be punished. In the following, I will not discuss this distinction between positive and negative versions of retributivism. The important thing is that the argument presented here is equally relevant for those who subscribe to a negative retributivist interpretation of proportionality.
If the algorithm is introduced, there will be some offenders (at the right side of μ in Fig. 3) who will be punished less severely than they would have been had the algorithm not been introduced. However, there will also be other offenders (at the left side of μ in Fig. 3) who will be punished more severely than they would have been had it not been introduced. Thus, one distribution does not seem preferable to the other. It might perhaps be suggested that there would be a difference if the disvalue of over-punishment increases in such a way that one unit of over-punishment (say, a day in prison) is morally worse the more severely an offender is being over-punished. However, it is hard to see how this view can be buttressed and worked out in theoretical detail, and no-one—to my knowledge—has ever defended such a view.
Obviously, there are major differences between the penal level in different countries. For instance, in the US a great number of crimes are punished much more harshly than in many other Western countries. However, if von Hirsch’s recommendation is taken for granted, then it becomes pretty clear that even those countries that use imprisonment more sparingly will be punishing more severely than what is ideally desirable.
Some retributivists may prefer to say that a fully developed theory of punishment—that is, one that provides the answer to how specific crimes should be punished—need not be based on absolute proportionality if this implies that it is only considerations of justice that provide this complete theory. It might be held that other types of consideration—say consequentialist consideration—will have to be involved within a proportionalist scheme in order to reach a complete theory. However, the important thing here is that theorists would subscribe to the view that a complete theory should provide answers as to how specific crimes should be punished (and that this is so, independently of whether one talks of absolute proportionality or simple punishments that are morally appropriate for particular crimes).
Needless to say, this is not tantamount to holding that reduced disparity would always be wrong.
References
Ashworth, A. 1998. Four Techniques for Reducing Sentencing Disparity. In Principled Sentencing, eds. A. von Hirsch & A. Ashworth, 227–239, Oxford: Hart Publishing.
Berk, R., et al. 2018. Fairness in Criminal Justice Risk Assessments: The State of the Art. Sociological Methods and Research 50 (1): 3–44.
Chiao, V. 2018. Predicting Proportionality: The Case for Algorithmic Sentencing. Criminal Justice Ethics 37 (3): 238–261.
Davis, M. 1992. Making the Punishment Fit the Crime. Boulder: Westview Press.
Divine, J.M. 2018. Booker Disparity and Data-Driven Sentencing. Hastings Law Review 69 (3): 771–834.
Duus-Otterström, G. 2019. Weighing Relative and Absolute Proportionality in Punishment. In On One-eyed and Toothless Miscreants: Making the Punishment Fit the Crime?, ed. M. Tonry, 30–50. New York: Oxford University Press.
Farmer, C., et al. 2017. Sentencing Inconsistencies: A Case Study. Australian Law Journal Reports 92: 517–528.
Frankel, M. E., 1972. Lawlessness in Sentencing, Cincinati Law Review 41: 1-43.
Husak, D. 2019. Why Philosophers (Including Retributivists) Should be Less Resistant to Risk-Based Sentencing. In Predictive Sentencing, ed. J. de Keijer et al. Oxford: Hart Publishing.
Kagan, S. 2012. The Geometry of Desert. Oxford: Oxford University Press.
Kehl, D., et al. 2017. Algorithms in the Criminal Justice System: Assessing the Use of Risk Assessments in Sentencing. Responsive Communities. https://cyber.harvard.edu/publications/2017/07/Algorithms.
Kleinig, J. 1973. Punishment and Desert. The Hague: Martinus Nijhoff.
Kopf, R.G. 2012. Judge-specific Sentencing Data for the District of Nebraska. Federal Sentencing Report 25: 50–52.
Lippke, R. 2012. Anchoring the Sentencing Scale: A Modest Proposal. Theoretical Criminology 16: 463–480.
Mason, C., and D. Bjerk. 2013. Inter-judge Sentencing Disparity on the Federal Bench: An Examination of Drug Smuggling Cases in the Southern District of California. Federal Sentencing Report 25: 190–193.
Mittelstadt, B.D., et al. 2016. The Ethics of Algorithms: Mapping the Debate. Big Data and Society 16: 1–21.
Murphy, J.G. 1979. Retribution, Justice, and Therapy. Dordrecht: Kluwer.
Roth, A. 2016. Trial by Machine. The Georgetown Law Journal 104: 1245–1305.
Ryberg, J. 2004. The Ethics of Proportionate Punishment. A Critical Investigation. Dordrecht: Kluwer.
Ryberg, J. 2010. Punishment and the Measurement of Severity. In Punishment and Ethics. New Waves, ed. J. Ryberg and A. Corlett. Basingstoke: Palgrave Macmillan.
Ryberg, J. 2019. Risk and Retribution: On the Possibility of Reconciling Considerations of Dangerousness and Desert. In Predictive Sentencing, ed. J. de Keijer et al. Oxford: Hart Publishing.
Ryberg, J., and J. Roberts, eds. 2021. Sentencing and Artificial Intelligence. New York: Oxford University Press (forthcoming).
Scheid, D.E. 1997. Constructing a Theory of Punishment, Desert, and the Distribution of Punishment? The Canadian Journal of Law and Jurisprudence 10: 441–506.
Schild, U.J. 1998. Criminal Sentencing and Intelligent Decision Support. Artificial Intelligence and Law 6: 151–202.
Scott, M. 2010. Inter-judge Sentencing Disparity After Booker: A First Look. Stanford Law Review 63: 1–66.
Simmons, R. 2018. Big Data, Machine Judges, and the Legitimacy of the Criminal Justice System. University of California Davis Law Review 52 (2): 1067–1118.
Singer, M. 1979. Just Deserts. Cambridge: Ballenger Publishing Company.
Stobbs, N., et al. 2017. Can Sentencing be Enhanced by the Use of Ethical Intelligence? Criminal Law Journal 41 (5): 261–277.
Tonry, M. 2019. Of One-eyed and Toothless Miscreants: Making the Punishment Fit the Crime?. New York: Oxford University Press.
von Hirsch, A. 1993. Censure and Sanctions. Oxford: Clarendon Press.
von Hirsch, A., and A. Ashworth. 2005. Proportionate Sentencing. Oxford: Oxford University Press.
Zerilli, J., et al. 2018. Transparency in Algorithmic and Human Decision-Making: Is There a Double Standard? Philosophy and Technology 32 (4): 661–683.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Ryberg, J. Sentencing Disparity and Artificial Intelligence. J Value Inquiry 57, 447–462 (2023). https://doi.org/10.1007/s10790-021-09835-9
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10790-021-09835-9