'If you profile, you will regret it; if you do not profile, you will also regret it': A review of profiling of the unemployed through statistics and artificial intelligence and the lifecycle of profiling models

Publikation: KonferencebidragPaperForskning


The use of advanced statistics and prediction algorithms has been systematically examined in other public sector areas, such as social welfare and social work (Gillingham and Graham, 2017), policing (Eubanks, 2018), and criminology (Lyon, 2006). However, the literature that deals with these technologies in the context of unemployment and public employment services is fragmented and lacks a scholarly foundation. This gap in the literature is somewhat contradictory to the scale of profiling, which can have severe consequences on those profiled, and that occurs in the majority of Western countries.

Therefore, this article conducts a systematic literature review of the literature on the subject of quantitative, statistical, and algorithmic profiling of the unemployed. This paper examines the research question: If statistical and algorithmic profiling of the unemployed can be understood through a phase-perspective? This review aims to provide a clear overview and to structure the existing literature. The overall ambition is to contribute to establishing a research field that is concerned with how advanced statistics and prediction algorithms affect the unemployed and impacts public employment services.

Based on the reviewed literature, this paper accounts for its descriptive findings and highlights specific gaps in the literature. This paper then develops a framework that conceptualizes the life cycle of quantitative profiling models. This framework consists of five phases and specific policy choices and inherent trade-offs to each phase. This conceptualization of profiling models suggests constructive ways to bridge the different scholarly positions identified in the literature. Ideally, the review charts the directions of future research, as further use of advanced statistics and prediction algorithms to profile the unemployed is likely (Desiere et al. 2019).
Publikationsdato24 aug. 2020
StatusAfsendt - 24 aug. 2020

Bibliografisk note

Kontakt mig gerne på kbhaug@ruc.dk, hvis du vil læse det :-)

Citer dette