Epistemic Rigor

Patterns & examples

Four epistemic states (in order of confidence):

1. Fact - Tested, verified, reproducible (high confidence)

   • "Go's defer runs in LIFO order" → write one test, it passes always

2. Test - Observed empirically but not fully verified (medium-high confidence)

   • "Pagination breaks on large datasets" → reproduced locally, haven't tested at scale

3. Assumption - Logical inference, not yet tested (medium confidence)

   • "User IDs are always positive integers" → sounds reasonable but unverified

4. Belief - Plausible but untested, may be wrong (low confidence)

   • "Database queries are probably the bottleneck" → intuition, no profiling yet

Pattern: Decision checklist

Before deciding, check your epistemic state:

Decision: Migrate to Firestore
Claim 1: "Firestore is cheaper than PostgreSQL" 
  → Belief (assumption based on marketing, not tested with our data size)
  → Action: Research pricing calculator with real numbers

Claim 2: "Migration will take 2 weeks"
  → Assumption (based on scope estimation, unverified)
  → Action: Build small spike to test one data type migration

Claim 3: "We need to migrate this year"
  → Fact? Assumption? → Check business requirements (might be belief based on false urgency)

Conclusion: Not ready to decide yet. Need (1) pricing analysis, (2) spike proof, (3) requirements clarification

Pattern: Debugging with rigour

Bug: Orders fail to save (belief: database issue)
Testing:
  1. Can we connect to DB? → Yes (test passes) → fact
  2. Can we insert a row manually? → Yes → fact  
  3. Can we insert via app? → No → narrows to app layer
  4. Does insert statement have correct syntax? → Build test case → fact
  5. Is transaction rolling back silently? → Add logging → fact

Result: Discovered silent rollback on constraint violation (fact) 
NOT database issue (was belief)

Anti-patterns to avoid

• ❌ Treating beliefs as facts (dangerous in decision-making)
• ❌ Skipping verification because something "feels right"
• ❌ Assuming you've tested something when you haven't
• ❌ Forgetting to update beliefs when evidence contradicts them
• ❌ Acting with 100% confidence when you have 40% certainty

KB Reference

~/vaults/baphled/3. Resources/Knowledge Base/AI Development System/Skills/Thinking-Analysis/Epistemic Rigour.md

Related skills

• critical-thinking - Rigorously analyse information before trusting it
• pre-action - Clarify what you know/don't know before deciding
• prove-correctness - Write tests to convert beliefs → facts