Reconstruct the information environment from records, not memory, before evaluating past decisions
Before evaluating any past decision, reconstruct the information environment that existed at decision time using contemporaneous records rather than memory, then evaluate the decision against that environment only.
Why This Is a Rule
Memory-based reconstruction of past information environments is unreliable because hindsight bias, outcome knowledge, and memory reconstruction all contaminate the reconstruction. You "remember" having more information than you actually had (hindsight), you "remember" warning signs that you noticed (but actually didn't notice until after the outcome), and you "remember" your reasoning as more complete than it was (reconstruction).
Contemporaneous records — emails, Slack messages, meeting notes, decision documents, git commits — provide the actual information environment: what was known, when it was known, and what was communicated. These records are immune to hindsight bias because they were created before the outcome was known.
The rule requires evaluating the decision against the reconstructed environment only — not against what you know now, not against what the outcome revealed, but against what was knowable at decision time. This is the only evaluation that produces genuine learning about decision quality rather than hindsight storytelling.
When This Fires
- During decision reviews, retrospectives, or post-mortems
- When evaluating whether a past decision was "good" or "bad"
- When someone is being blamed for a decision that produced a bad outcome
- Any retrospective evaluation where outcome knowledge might contaminate reasoning assessment
Common Failure Mode
Reconstructing from memory rather than records: "I remember we knew about the risk." Check the records — was the risk documented at decision time? Memory says yes (because you know the outcome); records might say no. Records are the ground truth; memory is the reconstructed narrative.
The Protocol
Before evaluating any past decision: (1) Gather contemporaneous records from the decision period: emails, messages, documents, meeting notes, decision records. (2) Reconstruct the information environment: what was known, what was uncertain, what was unknowable at that time? (3) Evaluate: given only the information available at decision time, was the reasoning sound? Was the confidence appropriate? Were alternatives adequately considered? (4) Only after this evaluation: introduce outcome information. The gap between "good reasoning given available information" and "bad outcome" reveals luck, not poor judgment.