The Compassionate Computer: Algorithms, Sentencing, and Mercy

In: Principled Sentencing and Artificial Intelligence (working title) J.V. Roberts & J. Ryberg eds., OUP, ‎Forthcoming

33 Pages Posted: 4 Nov 2020

See all articles by Netanel Dagan

Netanel Dagan

Hebrew University of Jerusalem - Faculty of Law

Date Written: September 14, 2020

Abstract

Sentencing scholarship largely neglects the possibility of applying algorithms to mercy. This ‎doesn’t come as a surprise: Is there any greater contradiction than between algorithmic decision-‎making and the compassionate, human and interpersonal nature of mercy? Such polarity brings ‎some theorists and policy makers to reject algorithm-based sentencing altogether. In this chapter, ‎we offer a preliminary attempt at integrating mercy within algorithmic sentencing. First, we ‎distinguish between two main concepts of mercy – justice and pure – and different types of ‎algorithms – deductive and inductive. Second, we argue: (a) As long as justice mercy can be ‎reduced to a proportionality-related calculus (e.g., extra harsh suffering) it can be introduced ‎through a deductive algorithm; (b) Pure mercy, being unpredictable, and deviating from justice, ‎can be incorporated mainly through an inductive algorithm. This is true, at least to some extent, ‎even for theories that permit deviation from equality when exercising mercy.‎

Keywords: sentencing, mercy, algorithms, proportionality, compassionate release, inductive/deductive ‎algorithms

Suggested Citation

Dagan, Netanel, The Compassionate Computer: Algorithms, Sentencing, and Mercy (September 14, 2020). In: Principled Sentencing and Artificial Intelligence (working title) J.V. Roberts & J. Ryberg eds., OUP, ‎Forthcoming , Available at SSRN: https://ssrn.com/abstract=3692449 or http://dx.doi.org/10.2139/ssrn.3692449

Netanel Dagan (Contact Author)

Hebrew University of Jerusalem - Faculty of Law ( email )

Mount Scopus
Mount Scopus, IL 91905
Israel

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
36
Abstract Views
228
PlumX Metrics