John Zerilli. 2020. “Algorithmic Sentencing: Drawing Lessons from Human Factors Research,” in Principled Sentencing and Artificial Intelligence, ed. J. Ryberg, J. Roberts & J. de Keijser. New York: OUP. - 1 - Algorithmic Sentencing: Drawing Lessons from Human Factors Research John Zerilli Abstract: Researchers in the field of “human factors” have long been aware that when humans devolve certain of their functions to technology, the transfer from human to machine can restructure more than the division of labour between them: humans’ perceptions of themselves and their abilities may also change. In particular, if a system becomes reliable enough, humans will become diffident to the point of adhering to the system’s recommendations even when they have grounds to disbelieve them. Such findings are relevant to the use of algorithmic and data-driven technologies, but whether they hold up in the specific context of recidivism risk assessment is only beginning to be considered. In this chapter, I describe and analyse some pertinent human factors results, and assess the extent to which they pose a problem for the use of algorithms in the sentencing of offenders. While the findings from human factors research are themselves robust, they do not seem to translate neatly to the judicial sphere. The incentives, objectives, and ideologies of sentencing judges appear to upset the usual pattern of results seen in many other domains of human factors research. Keywords: automation bias, automation complacency, human factors, human-computer interaction, sentencing