A Mulching Proposal: Satirising Algorithmic Ethics
A satirical paper by Os Keyes, Jevan Hutson and Meredith Durbin, written in the register of a straight-faced ethics audit and named after Jonathan Swift’s A Modest Proposal.
The setup: take the Fairness, Accountability and Transparency framework — the dominant toolkit for making algorithmic systems ethical — and apply it, rigorously and in good faith, to an algorithmic system for converting elderly people into nutrient slurry. The audit proceeds exactly as the methodology prescribes. The system’s FAT adherence improves substantially. It ends up demonstrably fairer, more accountable and more transparent.
The argument
That’s the whole point. If a framework can certify a monstrous system as ethical without ever asking whether the system should exist, the framework is measuring compliance, not ethics. The paper is a critique of procedural ethics-washing: audits, bias metrics and transparency reports optimise the how while leaving the whether entirely untouched, and an industry that adopts them can look rigorous while never confronting the underlying question.
Why it’s here
It has become a reference point in HCI and AI ethics circles, and it lands differently if you build things. Anyone shipping a model behind a fairness checklist, or filling out an impact assessment for a system whose premise nobody has interrogated, is the intended audience. The joke is load-bearing — the deadpan tone is what makes the gap between “passed the audit” and “is defensible” impossible to look away from.
Open access on arXiv (cs.CY).