Landscape Summary: Bias in Algorithmic Decision-Making: What is bias in algorithmic decision-making, how can we identify it, and how can we mitigate it?

Michael Rovatsos, Brent Mittelstadt, Ansgar Koene

Research output: Book/ReportCommissioned report

Abstract

As our societies become increasingly dependent on algorithms, so we are seeing our age-old prejudices, biases and implicit assumptions reflected back at us in digital form. But the algorithmic systems we use also have the potential to amplify, accentuate and systemise our biases on an unprecedented scale, all while presenting the appearance of objective, neutral arbiters.

This Landscape Summary draws together the literature and debates around algorithmic bias, the methods and strategies which may help to mitigate its impact, and explores four sectors in which this phenomenon is already starting to have real world consequences—financial services, local government, crime and justice, and recruitment. We identify a case study for each sector which may have significant consequences for individuals and groups in the UK: algorithmic loan redlining, child welfare, offender risk assessments, and CV sifting.

Through these case studies we have identified a number of key findings which are relevant for policymakers, regulators and other officials as they try to understand the socio-economic effects of algorithmic decision-making systems.
Original languageEnglish
PublisherUK Government
Number of pages70
Publication statusPublished - 19 Jul 2019

Fingerprint Dive into the research topics of 'Landscape Summary: Bias in Algorithmic Decision-Making: What is bias in algorithmic decision-making, how can we identify it, and how can we mitigate it?'. Together they form a unique fingerprint.

Cite this