Use before any research activity or significant decision. Reviews cognitive biases relevant to the current stage.
Pre-research and pre-decision bias review.
Identify context: What stage (L0-L5)? What activity (research, decision, evaluation)?
Load stage-specific checklist from cognitive-biases.md for the current scale.
For each bias on the checklist:
Check agent's own biases (from cognitive-biases.md agent section):
Output bias briefing:
## Bias Check: [Activity]
Stage: [L0-L5]
### Active Risks
| Bias | Risk Level | Mitigation |
|------|-----------|------------|
| ... | High/Med/Low | ... |
### Agent Self-Check
- Sycophancy: [OK/Risk]
- Recency: [OK/Risk]
- Pattern matching: [OK/Risk]
- Completionism: [OK/Risk]
### Proceed? [Yes / Yes with caution / Pause and address]
System-level bias diagnosis (Meza): Before attributing resistance or poor outcomes to cognitive bias, check if the "bias" is actually a rational response to a badly designed system. Ask:
| Phase | System Question | If Yes: Fix the System, Not the Person |
|---|---|---|
| Awareness | Does the person know the desired behavior? | Fix communication, not cognition |
| Motivation | Does the system reward the desired behavior? | Fix incentives, not mindset |
| Ability | Can the person realistically do it? | Fix resources/process, not training |
| Reinforcement | Is the behavior sustained by the environment? | Fix feedback loops, not habits |
| Sustainability | Can it persist without ongoing intervention? | Fix structure, not willpower |
If 2+ phases point to systemic causes, the primary intervention is system redesign, not individual debiasing. Update the bias briefing output above to flag the system-level root cause.
The 5 phases above are a systemic diagnostic inspired by behavioral design models (COM-B, Fogg Behavior Model), not a named framework. They are prompts for investigation, not rigid steps.
Source: Robert Meza (The Bias Gap — diagnostic tool, aimforbehavior.com). Not to be confused with Thaler & Sunstein's "Nudge" theory.
This skill should be triggered:
/diamond-progress or /diamond-assess detects bias-related gate failuresRecord bias check in canvas/opportunities.yml under bias_checks section with date, biases mitigated, and research design adjustments made.