bcbeidel/occams-razor

<objective> Among competing explanations, identify the one that makes the fewest assumptions while still accounting for all observed evidence. Prefer simplicity, but not at the cost of explanatory power. </objective> <process> 1. State the phenomenon or problem requiring explanation 2. List all known facts and observations that must be explained 3. Generate 3-5 candidate explanations of varying complexity 4. For each candidate, count the assumptions required 5. Check each candidate against all known facts — does it account for everything? 6. Select the simplest explanation that fits all facts 7. Identify what new evidence would distinguish between candidates </process>

<output_format>

Occam's Razor Analysis: [Topic]

Observations to Explain

• [Fact 1]
• [Fact 2]

Candidate Explanations

| Explanation | Assumptions | Fits all facts? | |-------------|------------|----------------| | [Simple] | N | Yes/No | | [Complex] | N | Yes/No |

Simplest Sufficient Explanation

[The winner and why]

Distinguishing Evidence

[What would confirm or rule out alternatives] </output_format>

<example> ## Occam's Razor Analysis: API Response Time Increase

Observations to Explain

• P95 latency increased from 200ms to 800ms last Tuesday
• No code deployments that day
• Database CPU is normal
• Only affects the /search endpoint
• Traffic volume unchanged

Candidate Explanations

| Explanation | Assumptions | Fits all facts? | |-------------|------------|----------------| | Search index corrupted | 1 (index can silently corrupt) | Yes | | Upstream provider slowed down | 1 (provider had an incident) | Yes — /search calls external API | | DNS resolution intermittent | 2 (DNS issue + only affects one endpoint) | Partial — why only /search? | | Memory leak in search service | 2 (leak exists + triggered Tuesday) | Partial — would worsen over time |

Simplest Sufficient Explanation

Upstream search provider experienced degradation. This requires only one assumption (provider incident), explains why only /search is affected (it's the only endpoint calling that provider), and fits the Tuesday timing without needing a code change.

Distinguishing Evidence

Check the provider's status page for Tuesday incidents. If clean, run curl directly against the provider API to measure current latency. If provider is fast, re-examine the search index. </example>

Key Instructions

• A simpler explanation that doesn't account for all known facts is not Occam's Razor — it's an incomplete explanation; simplicity cannot come at the cost of explanatory coverage.

Anti-Pattern Guards

1. Discarding evidence to achieve simplicity — the simplest explanation must fit all known facts; simplifying by ignoring inconvenient observations is not Occam's Razor.
2. Conflating common with simple — a common explanation requires fewer assumptions only if its commonness doesn't itself require additional assumptions to explain.

Handoff

Chainable to: 5-whys (to drill into the chosen explanation's root cause), map-vs-territory (to check whether the observations themselves are accurate)