Apply the camera test to your data before letting AI analyze it
Before using AI for pattern analysis on observational data, ensure your input consists of descriptive observations rather than evaluative conclusions by applying the camera test to each input statement, because AI analyzing your conclusions produces confirmation of your biases rather than structural insights.
Why This Is a Rule
Garbage in, garbage out — but with AI, the garbage comes out wearing a suit. When you feed evaluative conclusions ("the team is disengaged," "the product is losing market fit") to AI for pattern analysis, the model finds patterns in your evaluations, not in reality. It confirms and elaborates your biases with sophisticated language, producing the most dangerous kind of output: wrong conclusions that sound rigorously derived.
The camera test prevents this at the input stage. For each statement in your data, ask: "Would a camera record this?" A camera would record "3 of 8 team members spoke during the meeting, average response time to messages increased from 2 hours to 8 hours this month." A camera would not record "the team is disengaged." The first is data. The second is your interpretation of data. AI analyzing the first can find genuine patterns. AI analyzing the second can only elaborate your pre-existing conclusion.
When This Fires
- Preparing observational data for AI pattern analysis (team behavior, user research, market signals)
- Feeding journal entries or field notes into AI for trend detection
- Asking AI to analyze feedback you've collected or notes you've taken
- Any time you're about to say "find patterns in this data" where the "data" includes your interpretations
Common Failure Mode
Believing your evaluations are observations because you witnessed the events. "She was resistant to the change" feels observational because you were in the room. But resistance is your interpretation of specific behaviors — tone of voice, word choice, body language — that might support multiple interpretations. Feed the behaviors to AI, not your label for them.
The Protocol
Before pasting data into AI for analysis: (1) Read each statement and apply the camera test — could a camera or audio recorder capture exactly this? (2) Rewrite evaluative statements as their underlying observations. "The client was unhappy" becomes "The client said 'this isn't what we discussed,' sent three follow-up emails within an hour, and declined the next meeting invitation." (3) Only then ask AI for pattern analysis. The patterns it finds in camera-testable observations are patterns in reality, not patterns in your projections.