High confidence is a trigger for more scrutiny, not less — 8/10 means search harder
When confidence in a technical conclusion exceeds 8/10, treat that high confidence as a trigger to increase scrutiny and deliberately search for disconfirming evidence rather than reducing verification effort.
Why This Is a Rule
Rollwage et al. (2020) found that high confidence creates a neural gate that blocks disconfirming evidence from being processed. When confidence exceeds a threshold, the brain literally stops updating based on contradictory information. You feel increasingly certain not because the evidence is stronger, but because your brain is filtering out the evidence that would weaken your certainty.
This makes high confidence the most dangerous state for technical conclusions. The natural response to high confidence is to reduce verification effort: "I'm 90% sure it's the database, so I don't need to check the network." But 90% confidence combined with the neural gating effect means you're most likely to miss the evidence that it's actually the network — precisely because your high confidence prevented you from seeing it.
The rule inverts the natural response: high confidence → more scrutiny, not less. When confidence exceeds 8/10, it should trigger the same alarm as a failed test — something requires closer examination, not less.
When This Fires
- When you feel very confident about a bug's root cause before thorough investigation
- When a technical architecture decision feels "obvious" and doesn't seem to need review
- When you're about to skip verification because you're "sure" about the answer
- Any technical conclusion where confidence arrives before comprehensive evidence
Common Failure Mode
Using high confidence to justify skipping steps: "I'm pretty sure this is right, so I'll skip the load test." The higher your confidence, the more dangerous the skip — because the neural gating effect means you're least likely to notice the evidence that would change your mind.
The Protocol
When confidence in a technical conclusion exceeds 8/10: (1) Notice the confidence level — rate it explicitly. (2) Treat it as a trigger: "My confidence is high, which means my evidence-filter is narrowing. I should search harder, not less." (3) Deliberately seek disconfirming evidence: "What would make this wrong?" (4) Maintain or increase verification effort — the opposite of what confidence suggests. (5) If the conclusion survives active falsification attempts, your high confidence is earned. If it doesn't, you've caught an error that high confidence would have hidden.
Source Lessons
Confirmation bias operates in real time
You unconsciously seek and emphasize evidence that confirms your existing beliefs.
When in doubt, wait
When you cannot distinguish signal from noise, the highest-value action is usually inaction. Time is a filter — it degrades noise and amplifies signal. Forcing a decision under ambiguity does not resolve uncertainty; it converts uncertainty into error.