Recent debate has centred on the usage of algorithms in society. Such questions become more immediate when they impact our daily lives such our health, or our justice system. Together with UCL Health Informatics researchers Julie George and Henry Potts, Statistical Science staff Sofia Olhede and Patrick Wolfe (also at Purdue) are organising a meeting to question how we approach such problems.
Algorithms are increasingly used to make decisions affecting our lives, but can we ensure machine-generated decisions are fair? Algorithms have long been used in the private sector to make marketing or finance decisions. As large-scale data becomes available in the public sector, these techniques are increasingly being deployed in areas as diverse as education, criminal justice or health. If these algorithms are based on data which misrepresents specific groups of people, then their use could widen social inequalities or breach equality laws.
By developing our understanding of how algorithms can be unfair and what their impact in different areas of the public sector, we can contribute to the responsible development and use of these decision aids. Prior work has focussed on algorithms directly affecting individuals, but big data is also used to allocate resources to different groups within the population. Do decisions to allocate resources informed by predictive tools unfairly decrease access to resources?