Ontological Hygiene

Abstract

This paper argues that some of our most important concepts should be treated like critical infrastructure. When labels such as “risk,” “security,” or “deserving” are built into laws, databases, and technical systems, they can quietly help decide who is watched, excluded, or harmed. Drawing on ideas from conceptual engineering, conceptual ethics, and computer security, the paper introduces ontological hygiene: a practice of systematically stress-testing high-impact categories before and after they are deployed. Borrowing the notion of “0-day” vulnerabilities from cybersecurity, it proposes “-1-day testing” for concepts: adversarially probing likely misuses and failure modes in advance, rather than only reacting once harm has occurred.

The core contribution is an Ontological Hygiene Meta-Test, a short set of questions for evaluating high-impact concepts along dimensions of impact, misuse scenarios, scope and boundaries, power dynamics, and conditions for revision or retirement. The paper also sketches the role of conceptual red-teamers who deliberately “attack” proposed categories, and argues for better documentation and reporting channels so that conceptual failures are logged and learned from, rather than rediscovered in each generation. Historical cases, including dynamite and the Maxim gun, illustrate how reassuring frames like “safety,” “efficiency,” and “progress” can mask predictable harms when they are not examined under real social and political conditions. The paper concludes by outlining what a more justified ontology would look like: a working vocabulary in which high-impact concepts are explicitly designed, documented, periodically re-evaluated, and treated as objects of ongoing safety work rather than invisible background assumptions.

This manuscript is under consideration at the Carolina Undergraduate Research Journal. Comments welcome.

Made in collaboration with ChatGPT 5.1 Thinking.