Principles

How I approach claims

• Make assumptions explicit. If a conclusion depends on an assumption, it should be visible.
• Prefer designs over algorithms. Strong identification beats fancy models.
• Show sensitivity. If small changes flip results, that’s a result.
• Reproducibility by default. Clean structure, pinned dependencies, and runnable outputs.

What “trustworthy” means here

• Clear scope: what a method can and cannot answer
• Sources when factual claims matter
• Honest uncertainty when evidence is thin