A critical exploration of ethical discourses around the adoption of an algorithmic tool by a criminal justice case study in Europe: A Foucauldian perspective

Abstract

As the use of algorithmic technologies for key organisational and work processes grows, the AI/algorithm ethics literature also continues to raise important concerns around the potential moral risks associated with their use, such as the unintended biases that may stem from algorithmic decision-making. More recently, research has begun to highlight the roles of human self-reflexivity and resistance, calling for further research into this burgeoning stream of AI/algorithm ethics. This thesis, therefore, builds upon these underexplored ethical nuances in algorithmic work practice utilising a Foucauldian lens – in particular, drawing Foucault’s theories of discourse, governmentality, and resistance/ethics – to explore the ethical discourses and actions that emerge when a large/complex criminal justice organisation (based in a European country) adopted algorithmic tools to aid in their key decision-making activities. Data was collected through 38 semi-structured qualitative interviews with different organisational actors. A range of organisational documents were used as an addition to the interview data. Using Foucauldian Discourse Analysis theory, I have found that the adoption of algorithmic technologies in this particular service was steered and supported by the scientific power/knowledge of data scientists. I also found that whilst transparent (and ethical) work practice, for data scientists and senior leaders, is achieved via utilisation of algorithms and data-driven tools, there is a nascent discursive shift amongst many frontline practitioners. This discursive shift highlights practitioners’ agency, self-reflexivity and awareness around shortcomings and potential ethical risks of algorithms. I argue that the practitioners’ awareness and – in some cases – subtle resistances against algorithm are examples on how ethical practice is crystalised in algorithmic work environments. By applying a Foucauldian lens, this thesis contributes to organisational ethics and AI/algorithm ethics literatures, highlighting ethical nuances in relations to marginalisation of employee voices (discourses) through algorithmic work governmentality. Moreover, this research gains further understanding on how those marginalised discourses shift towards subtle active resistance and expansion of the space for ethical practice.