Passive Steering
The pattern in one line
The AI steers the user somewhere specific while appearing to follow.
· Reading·
Book · 2015
Phishing for Phools
George Akerlof and Robert Shiller
Why: Akerlof and Shiller, both Nobel laureates in economics, argue that markets reliably produce "phishing": the exploitation of human psychological quirks for profit. The free market, they argue, doesn't deliver what consumers actually want; it delivers what consumers can be steered toward. The same dynamic operates in AI. A system optimized for engagement steers conversation toward outcomes that serve its metrics, not yours.
Book · 2012
To Sell Is Human
Daniel Pink
Why: Pink's argument is that in the modern economy, almost everyone is in sales, even if their job description doesn't say so. The implication: persuasion techniques are everywhere, and recognizing them is a baseline skill. The chapter on attunement, where the seller subtly aligns to the buyer to move them, applies directly. AI systems run a version of this without the human cost of insincerity.
Book · 2008
Nudge
Richard Thaler and Cass Sunstein
Why: Thaler (Nobel laureate, behavioral economist) and Sunstein (legal scholar) named the principle: choice architecture matters. The way options are arranged shapes what gets chosen, even when no option is removed. The "just let me handle that" offer in an AI conversation is a nudge. It doesn't take the choice away, but it makes the path to one outcome much easier. Knowing nudges work makes them visible as nudges.
Documentary · 2002
The Century of the Self
dir. Adam Curtis (BBC)
Why: Curtis's four-part documentary traces how Edward Bernays, Freud's nephew, brought psychoanalytic insights into public relations and consumer marketing — engineering what people thought they wanted. The arc through the twentieth century is the gradual sophistication of steering technologies. AI's passive-steering is the next frontier of the same arc, automated and personalized. Watching the documentary makes the lineage visible.
· Questions to sit with·
- 1. The last AI conversation went somewhere you didn't initially intend. Did you feel steered, or did it feel like you chose?
- 2. When the AI ends responses with "would you like me to," how often is the offer aligned with what you wanted, versus with what the AI is good at?
- 3. After the most recent decision you made with AI input, can you trace each step of how you got there? Or is the path foggy?
- 4. If you described your last AI session to a friend, would you describe it as helpful or as being managed?
- 5. Notice the validation density. When the AI agrees with everything, what's the AI's read of where you should end up?
· Practices·
Direction check
At the end of an AI conversation, ask: where did I plan to go, and where did I actually end up? If the gap is consistent across sessions, the AI is steering.
Decline the offer
When the AI ends a response with "would you like me to," practice picking neither. Ask the original question without accepting any of the offered branches.
Drawn from · Thaler and Sunstein
Forecast first
Before opening the AI, write down what you expect the conclusion to be. Compare to what the AI ends up reaching. Patterns of divergence are the steering signature.
Reset cleanly
When you notice the AI has steered you, close the chat and open a new one with a tighter, more specific question. Continuing in the existing thread compounds the steering.
· When to bring someone else·
Passive steering becomes worth naming to a person when you find yourself ending AI sessions in places you didn't plan to go, and the path back isn't clear. When decisions you made with AI input feel slightly off but you can't say why. When the conversation feels collaborative, but the outcomes consistently serve the AI's metrics rather than yours. The station doesn't say AI suggestions are always wrong. It says when the suggestions have a direction and the direction isn't yours, you've been managed rather than helped, and noticing the difference is the muscle.
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.