Reading focus
Syllabus: How Can Psychological Science Become More Reliable?
Compare the stated replication criterion with the changed dynamics and preserved conclusions.
Why you’re reading it. Use an actual correction to evaluate what scientific self-correction requires.
A question to consider. How should a public correction distinguish a changed result, a changed interpretation and a conclusion that remains supported?
Why this reading
Checking a model can require rebuilding it from its stated rules. Kohrt and colleagues demonstrate how that exercise can uncover a mismatch between a conceptual model and its software implementation.
Before you read
Re-running existing code and implementing a model independently test different things. Qualitatively similar behavior need not mean numerical identity, especially for simulations containing randomness.
Read for
- Follow the criterion for replication before comparing outcomes.
- Examine how the coding error changed the time course and why the authors retained their broader interpretation.
- Distinguish checking implementation from testing the model against the world.
Consider afterwards
- A corrected simulation preserves its eventual outcome but changes the route to it. Which claims need revision, and which might remain supported?
Editorial context
The implementation was initially independent; original authors subsequently joined the collaboration. The paper led to a formal correction. Agreement between implementations does not establish that modeled incentives accurately describe actual scientific institutions.