Automated Decisions, Human Judgement: Discretion in Algorithmic Public Administration

Public bodies increasingly decide who receives a benefit, a licence, an inspection or a place on a waiting list with the help of an automated system. The system is rarely the decision-maker in law. In practice it sets the order of the queue, flags what deserves attention, and supplies a number that a person is then asked to agree or disagree with — usually under time pressure, and usually knowing that disagreeing takes longer than agreeing.

This special issue asks what happens to professional discretion when it meets a system of this kind: where discretion narrows, where it survives, and where it moves somewhere less visible — into how a case is first classified, which fields get filled in, or which referrals quietly never get made.

What we are looking for

We welcome empirical, theoretical and methodological contributions from any discipline. Work that sits between disciplines is actively wanted; the journal exists for it. Topics include, but are not limited to:

• Street-level practice around risk scores, triage tools and eligibility engines
• How an automated recommendation is recorded, overridden, or deferred to in the written record — and what the record leaves out
• Procurement, pilots and the gap between what a system was bought to do and what it does on a Tuesday afternoon
• Accountability in practice: appeals, ombudsman cases, audits, inspection regimes
• The labour that keeps these systems running — data entry, exception handling, escalation, maintenance
• Comparative and cross-national studies, particularly outside the usual set of countries
• Methods for studying decisions that leave thin or misleading traces
• Participatory and co-design approaches with caseworkers and with the people subject to the decisions

What makes a strong submission

We are looking for work grounded in a specific setting and honest about its limits. A paper that explains carefully how twelve caseworkers in one agency actually use a tool is more useful to this issue than one that generalises about automation from nothing in particular.

Negative and null findings are welcome. So are papers that report a system working well; a literature that only documents failure is not describing the world.

Article types

We accept research articles, review articles, research notes and commentaries. Research notes are a good fit for a single well-documented case or a methodological problem that does not need a full-length treatment.

How to submit

Submit through the journal's own system and select this special issue when prompted. There is no separate portal and no fee at any stage: the journal is open access and charges authors nothing.

Submissions are reviewed anonymously, so please remove identifying details from the manuscript file itself. Author information is collected separately in the submission form. Everything published is released under a CC BY 4.0 licence, so authors keep the copyright in their work.

An expression of interest is optional but helpful — a title and a paragraph by 15 December 2026 lets us plan reviewer capacity, and it is not binding either way.

Timeline

Accepted papers are published online as soon as they are ready rather than waiting for the issue to close, so an early acceptance is not penalised with a long wait.

Questions

Write to editorial@sociotechnicalinquiry.org. If you are unsure whether a paper is in scope, send the abstract and ask — it takes us less time to answer than it takes you to guess.