Academic journey
Alexandra Morgan did not begin in computer science. She read philosophy at Edinburgh, where a seminar on the epistemology of testimony left her with a question she could not put down: what does it mean to trust a claim you cannot verify? Two decades later, that question has become a research programme spanning machine learning, administrative law and the practical business of auditing systems already in production.
Her route into the field was indirect. After a master's in logic she spent three years as a policy analyst, drafting briefings on automated decision-making for a legislature that had not yet decided whether the subject was technical or constitutional. The experience convinced her that the two vocabularies were not being translated for one another. Engineers were producing accuracy figures; regulators were asking questions about reasons. Nobody was building the bridge.
She returned to doctoral study late, at thirty-one, and completed a PhD on the formal semantics of model explanation. The thesis was unusual in that it contained both proofs and statutes. It argued that the prevailing family of post-hoc explanation methods could not, in principle, support the standard of reasoned decision required by administrative law — and then set out an alternative class of intrinsically interpretable architectures that could.
The argument was contested. It is now largely accepted. In the decade since, Morgan has led the Institutional Intelligence Group, a laboratory that is deliberately half technical and half legal, and which has been engaged by hospital trusts, benefits agencies and two national audit offices to examine systems already making consequential decisions about people.
She teaches a graduate seminar that requires computer scientists to read case law and lawyers to write code. It is heavily oversubscribed.
