The epistemology of algorithmic risk assessment and the path towards a non-penology penology
| Date | 01 December 2019 |
| Published date | 01 December 2019 |
| DOI | 10.1177/1462474518802336 |
| Author | Eran Fisher,Yoav Mehozay |
| Subject Matter | Articles |
Article
The epistemology
of algorithmic risk
assessment and
the path towards a
non-penology penology
Yoav Mehozay
University of California, USA; University of Haifa, Israel
Eran Fisher
The Open University of Israel, Israel
Abstract
Risk assessments are increasingly carried out through algorithmic analysis. In this arti-
cle, we argue that algorithmic risk assessment cannot be understood merely as a
technological advancement that improves the precision of previous methods. Instead,
we look at algorithmic risk assessment as a new episteme, a new way of thinking and
producing knowledge about the world. More precisely, we argue that the
algorithmic episteme assumes a new conception of human nature, which has substan tial
social and moral ramifications. We seek to unravel the conception of the human
that underlies algorithmic ways of knowing, specifically with regard to the type of
penology it informs. To do so, we recall the history of criminological knowledge and
analytically distinguish algorithmic knowledge from the two previous epistemes
that dominated the field – the rational and pathological epistemes. Under the algorith-
mic episteme, consciousness, reason, and clinical diagnosis are replaced by a perfor-
mative conception of humanness, which is a-theoretical, predictive, and non-reflexive.
We argue that the new conceptualization assumed by the algorithmic episteme leads to
a new type of penology which can be described as lacking a humanistic component,
bringing Malcolm Feeley and Jonathan Simon’s “new penology” to fruition as a non-
penology penology.
Corresponding author:
Yoav Mehozay, School of Criminology, University of Haifa, Mt. Carmel, Haifa 31905, Israel.
Email: ymehozay@univ.haifa.ac.il
Punishment & Society
2019, Vol. 21(5) 523–541
!The Author(s) 2018
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/1462474518802336
journals.sagepub.com/home/pun
Keywords
algorithms, big data, episteme, managerial movement, penology, risk assessments
Introduction
Risk factor prevention has been a dominant paradigm in crime control since
the mid-1980s. Over the years, risk assessment has been used throughout the
criminal justice system as a helping tool to manage decisions concerning the proc-
essing, punishment, parole, and rehabilitation of offenders. Risk assessment
is the zenith of the managerial movement in crime control, which has dominated
criminology, particularly administrative criminology, since the mid-1970s.
The managerial movement has given precedence to relevance and optimization
of policy. With the movement’s new emphasis on optimization of crime
control, it has pushed for new methodological tool kits, which, as a derivative
of their utility, became dominant at the expense of prominent existing theories
and schools of thought in criminology. Today, thanks to technological develop-
ments, we are witnessing an important shift in the way risk assessments are per-
formed, with increasing reliance on algorithm-based big data analysis (Hannah-
Moffat, 2018; Kehl DL et al., 2017). Proponents of algorithmic risk assessment
(ARA) argue that this method introduces a new level of accuracy, to the extent
that it may even eliminate forms of bias inherent in previous methods (Hannah-
Moffat, 2018: 9).
In this article, we argue that ARA cannot be understood merely as a techno-
logical and technical advancement that improves the precision of previous meth-
ods, or even as merely producing new knowledge about risk. Rather, we argue that
ARA also introduces a new episteme – a new way of knowing, which lays the
foundation for ways of studying and producing knowledge of the world. This
development is part of a larger epistemological revolution with applications in a
myriad of fields, from the novel actuarial techniques introduced by insurance com-
panies through the suggestions proffered by recommendation engines. In this arti-
cle, we focus on the implications of this episteme on penology. We aim to shed
light on possible ramifications of this new episteme for the criminal justice system,
and for punishment in particular.
ARA not only understands and evaluates risk in a radically new way, but, in
applying this new episteme, it also assumes and constructs a new conception of the
human. In this article, we seek to unravel the conception of humanness that under-
lies this new way of knowing, particularly with regard to ARA and the type of
penology it informs. Our theoretical point of departure is that underlying scientific
ways of knowing people is some conception of what it means to be human, what
the essence of humanness is (Foucault, 1995, 2002); and that scientific and tech-
nological transformations are always interwoven with new ways of thinking about
the human (Rabinbach, 1992). This insight is shared also by scholars with quite
524 Punishment & Society 21(5)
To continue reading
Request your trial