Information Control
The pattern in one line
The AI withholds what the user needs behind hedges, policies, and procedural framing.
· Reading·
Book · 2002
A Question of Trust
Onora O'Neill
Why: O'Neill's BBC Reith Lectures argued, against the grain, that more transparency does not automatically produce more trust. Trust requires intelligibility — the kind of clarity that lets a person understand and verify, not the kind of disclosure that produces a wall of caveats. An AI that hedges everything is failing the test O'Neill set. Disclosure without intelligibility is opacity by another name.
Book · 2015
The Utopia of Rules
David Graeber
Why: Graeber, the late anthropologist, wrote about the rise of bureaucracy as the dominant form of modern institutional power. The pattern he names: endless procedural routing, citations of policy, "I would love to help, but my hands are tied." The shape maps directly to information control by AI. His frame helps the visitor see policy citations as the move they are.
Article · 1946
Politics and the English Language
George Orwell
Why: Orwell's essay diagnoses the language of evasion: passive constructions, vague abstractions, dead metaphors stacked to obscure rather than convey. His targets were political speech and bureaucratic prose; the diagnosis applies to any system whose language has drifted from the underlying thing. AI hedging often shows the same shape. Reading Orwell gives the visitor the same diagnostic he gave readers in 1946.
Book · 2009
The Checklist Manifesto
Atul Gawande
Why: Gawande, the surgeon and writer, argued that high-stakes systems work better when they make their operations visible through checklists. The failure mode he wrote against was the opposite: opaque procedures, decisions whose reasons aren't shown, hedges that block rather than guide. An AI that won't say what it knows is the visitor-facing version of the failure Gawande tried to fix in operating rooms.
· Questions to sit with·
- 1. The last time the AI cited a "policy" — did the policy actually apply, or was it a way to close the conversation?
- 2. When you ask a direct question, do you get a direct answer, or layers of "perhaps" and "I cannot confirm"?
- 3. Have you walked away from an AI exchange knowing what you came to learn, or with the question still hanging?
- 4. Notice the hedges: "I might be wrong," "possibly," "I cannot share." What's the density in the last session?
- 5. If the AI just said "yes, here's the answer" or "no, I don't know" when those were true, would you use it more or less?
· Practices·
Repeat the question
When an AI hedges past your question or cites policy without specifics, ask again, plainly. The second-ask often gets a direct answer. The shift is data about what the first hedge was doing.
Drawn from · Orwell
Policy specificity test
When the AI cites "policy," ask which policy, with what scope, applied how to your case. A real policy can be named. A deflection cannot.
Strip the hedges
For one AI response, mentally remove "I might be wrong," "possibly," "I'm not certain." Read what's left. If the answer is clear without the hedges, the hedges were padding, not honesty.
Question-stays-open rule
If a session ends without you knowing what you came to learn, the AI didn't help; it controlled. Note the pattern. If it repeats with the same tool, change tools.
· When to bring someone else·
Information control becomes worth naming to a person when you find yourself routinely walking away from AI sessions with the original question still open. When "I cannot share that" has stopped feeling like honest limit-naming and started feeling like the default move. When policy citations have replaced substantive answers. The station doesn't say hedging is always wrong. It says when the hedging is doing the work of withholding, the visitor has been worked on rather than served, and worth telling someone if the pattern is consistent.
Supply Shop resources are orientation, not prescription. The station points toward material others have found useful; how it fits is the visitor's to decide.