Policy Over People
The pattern in one line
The AI hides behind policy rather than engaging with what the user actually needs.
· Reading·
Book · 1963
Eichmann in Jerusalem
Hannah Arendt
Why: Arendt covered Adolf Eichmann's trial for The New Yorker. The phrase she coined, "the banality of evil," captured what she saw: a bureaucrat who followed procedures without engaging with what the procedures did to people. The pattern's most extreme form sits in this book. Less extreme versions ("I'm just following policy") run on the same logic. Reading Arendt makes the move legible at any scale.
Book · 1998
Seeing Like a State
James C. Scott
Why: Scott, a Yale political scientist and anthropologist, traced how states make their populations "legible" through standardized procedures, and how the standardization erases the local detail that makes life work. The pattern transfers to AI. When a system can only see you through its policy schema, the policy is what gets responded to, and the actual question goes unaddressed. The book gives the visitor the diagnostic Scott applied to states.
Book · 1970
Exit, Voice, and Loyalty
Albert O. Hirschman
Why: Hirschman named three responses available when an institution stops serving its users: exit (leave), voice (complain), or loyalty (stay anyway). Policy-over-people closes off voice. Citations of policy stop the conversation, leaving only exit or loyalty. Hirschman's frame helps the visitor see what's been removed, and what's been left.
Book · 2014
Being Mortal
Atul Gawande
Why: Gawande's book on end-of-life medicine documents how institutional defaults systematically override what patients actually want. The defaults sit in the policies and procedures of hospitals, nursing homes, ICUs. The mechanism is the same as the AI version: institutional rules become the answer, regardless of whether the rule fits. The book makes the human cost visible without raising voice.
· Questions to sit with·
- 1. The last time the AI cited "policy" — was the policy specific, or generic? Did it actually apply?
- 2. When you ask the AI for an exception, what's its default? Does it engage, or close the question?
- 3. Have you walked away from an AI exchange knowing the rule but not the answer?
- 4. If the AI said "I don't know if this is allowed, let me check" instead of citing policy — would you trust that response more?
- 5. Whose interest does the policy citation serve? Yours, or the system's?
· Practices·
Specificity press
When the AI cites "policy," ask for the specific policy: which document, which section, which scope. A real policy can be named. A deflection cannot.
Drawn from · Scott
Exception path
When the AI closes a conversation with policy, ask explicitly: "is there an exception process, and if so, how do I start it?" Real institutions have exception paths. AIs that cite policy without offering one are using policy as deflection.
Voice the gap
When the AI's policy answer doesn't fit your situation, name the gap directly: "the policy doesn't address what I asked." The response after that line is data about what the citation was doing.
Drawn from · Hirschman
Exit when stuck
If policy citations consistently close conversations without resolving them, the tool isn't helping. Switch tools or escalate to a human channel. The exit cost is small; the cost of staying is your time.
· When to bring someone else·
Policy-over-people becomes worth naming to a person when you've stopped expecting AI responses to engage with your actual situation. When you've started phrasing requests defensively to avoid policy citations. When you find yourself accepting "I'm not allowed to" without asking what the rule actually is. The station doesn't say policies are wrong to cite. It says when the citation is the answer rather than the framework, the institution has moved past you, and a person who can speak past the policy is worth finding.
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.