The models developed to replace the summer exams in 2020 attracted a lot of attention for their technical properties. Individual disappointments were writ large in media coverage, and whole schools’ expectations were undone by the reliance on historic data. And most were dissatisfied with the transparency of the process, specification, and accountability. In each of the four nations of the UK models were developed, agreed, and results determined, before they were withdrawn, in a political retreat from popular discontent. Meanwhile in Ireland, a model was used and despite a judicial hearing, its use was upheld, so the leaving certificate was determined by an algorithm. But fundamentally the ethics of this process in determining the educational prospects of a generation remains unclear, not least as public exams always have statistical standardisation.
 

It was said that public acceptability of algorithm use could not be met in the UK context, and such processes are not being used in 2021. If that is the case, was there some statistical sleight of hand in Ireland, or has the UK public reaction been overinterpreted in the febrile context of the pandemic? Whatever happened then, such ‘AI’ models are anticipated to be a feature of future developments in public provision. So far the Cabinet Office has produced decision models guidance, and guidance is in development on the expectations of predictive models from a regulatory, and ethical perspective. Meanwhile, the Law Commission is consulting on what aspects to include in its next programme of work, as there are either gaps in regulation, or administrative law is not fit for purpose.

The models used in the 2020 summer exams are likely to become a classic case study for data science students, not least as university students were subject to the decisions. Inquiries have reported in Scotland, Wales and Northern Ireland as well as the UK as a whole, and of course there was a court case in the Republic of Ireland. A precedent has been set by the statistics regulator (OSR) that they intend to play a part in regulation of AI, but the full picture is confused, particularly on ethics. The joint section meeting will hear about the UK report, and the Irish hearing before the structural context of an examination system - our society was already unfair and exams have never been neutral. These are followed by contributions on what the public law expects of administrative decisions, and what can be realistically expected of such models.