Debrief difficult social interactions in 5 minutes — compare what your agent prescribed vs. what you actually did
After any difficult social interaction, complete a 5-minute written debrief comparing what your social agent prescribed versus what you actually did, documenting the gap to update the agent based on real behavioral data.
Why This Is a Rule
Social behavioral agents — rules for how to handle conflict, give feedback, or navigate difficult conversations — are uniquely difficult to evaluate because the context is always slightly different and emotional arousal distorts real-time self-observation. The 5-minute written debrief solves both problems: writing forces specificity ("I said X when the agent prescribed Y") and the temporal distance (after rather than during) allows the emotional arousal to subside enough for honest assessment.
The prescription-vs-reality comparison is the key diagnostic structure. Every gap between what the agent prescribed and what you actually did is a data point. If you consistently deviate at the same point — "My agent says to ask questions before defending, but I always start defending" — that's not random; it's a pattern revealing either that the agent's action step is unrealistic for you under emotional load, or that the trigger fires too late to interrupt your default.
Five minutes is the Goldilocks duration: enough to capture the meaningful gaps, short enough to actually complete consistently. Longer debriefs produce richer data but lower compliance; shorter ones miss the nuances that matter for agent updating.
When This Fires
- After any difficult social interaction where you had a relevant behavioral agent
- After conflicts, hard conversations, feedback sessions, or negotiations
- After interactions where you noticed yourself deviating from your intended approach
- During weekly reviews when processing accumulated debriefs for agent updates
Common Failure Mode
Debriefing only when it went badly: "I should review what went wrong." But reviewing successes is equally valuable — when you followed the agent's prescription and it worked, that's confirmation data. When you deviated from the prescription and it worked better, that's update data. Both outcomes inform agent improvement.
The Protocol
(1) Within 30 minutes of a difficult social interaction, write for 5 minutes. (2) Answer three questions: What did my agent prescribe for this situation? What did I actually do? Where and why did they differ? (3) For each gap, classify: was the agent's prescription unrealistic for this context? Was my default behavior actually better? Or did I fail to execute a prescription that would have been better? (4) Based on the classification: update the agent (prescription was wrong), note a win (you followed it and it worked), or note a training need (prescription was right but you couldn't execute it). (5) Accumulate debriefs and review weekly for patterns across interactions.