CONCEPT
AI Outputs as Plans, Not Actions
Suchman's sharpest diagnostic proposition: AI generates plans addressed to described situations, not actions tested against encountered ones — and the most dangerous institutional error is to confuse the two.
AI outputs are plans, not actions. This is Suchman's most clarifying proposition for thinking about large language models, algorithmic targeting, and every domain in which AI-generated results are deployed and trusted. A plan is a representation of action made in advance; an action is what actually occurs when an agent engages with specific circumstances. Plans address described situations — the prompt, the specification, the training corpus. Actions address encountered situations — the real deployment environment, the actual courtroom, the specific patient. The gap between described and encountered is permanent, and the catastrophic institutional error is treating plans as actions: accepting AI-generated outputs as if they had already been tested against the reality they claim to address.
In The You On AI Field Guide
The proposition extends Suchman's original distinction between plans and situated action to the specific case of AI-generated outputs. A plan is useful: it provides orientation, structure, a starting point. What a plan