Most Freedom of Information (FOI) decisions look inconsistent because we never see the logic underneath them. A black box on a page feels personal. Like the officer improvised. Like the agency had something to hide. The argument starts there, in the absence of a map.
I kept noticing this across Australian FOI matters (Freedom of Information Act 1982 (Cth) and the state equivalents). Different agencies, similar exemptions, but the outcomes never lined up. Sometimes s 47F swallowed whole paragraphs. Sometimes the same clause barely touched a document. It took a while to realise the mess wasn't random. It was structural.
FOI law has a shape. You just rarely get to see it.
At its core, every redaction turns on a few questions: what exemption is being relied on, what harm is being claimed, whether a public interest test applies, and how that test is actually applied in practice. None of this is visible when you're staring at black ink on a PDF.
So I built the Redaction Taxonomy as a way to surface the hidden logic and give practitioners, researchers and students a shared language for describing redaction patterns.
The taxonomy groups redactions into categories people already recognise, then adds labels for the failure modes that matter most: over-extension, mechanical invocation of exemptions, harm tests asserted without reasons, and decisions that lean on risk language without articulating the risk.
Version 0.2 adds something I wish I'd had sooner: a statute-aligned decision-logic visualiser that turns the FOI pathway into an explicit map. You can trace the route from trigger to exemption to any public interest step, and see where discretion sits. It also makes it obvious how two agencies can read the same words and still land in different places.
This is not a reform proposal. It's infrastructure.
When the logic is visible, the arguments change. You stop debating personalities. You start debating tests, thresholds and evidence.
If you want to read it, it's on glasscase.org.
Comments