Do not trust my or any computational research
summary
Argues that peer review routinely fails specific checks and offers reproducibility examples: a Python reproduction of a 2012 model cannot reach the paper's 97% reader‑review accuracy, wrapping Robert Axelrod's Fortran into the axelrod Python library shows the published tournament outputs do not match the available code, and widespread Excel gene-name conversions imply ~30% of genetic studies could be wrong. Proposes CI, test suites and publishers running authors' checks before acceptance.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 3:37Wittgenstein’s thought experiment
- 11:06Models and peer review
- 21:52Bacon’s scientific method
- 28:57Scientific process failures
- 31:52Reproducibility in research
- 39:58Future of peer review
- 46:41Q&A
related talks
29mClean Code for Good Science: Rust in Research and Health
34mFrom hard science to baseless opinions: where did we go wrong?
42mWhat our Brains can Teach us About Software Engineering
41mWhy Software Engineers Disagree About Everything
38mIntro to Empirical Software Engineering: What We Know We Don't Know
34mFrom Vibe Coding to Vericoding
43mLearning to Live with Errors
06mØ