Accuracy is not a model evaluation plan
A practical checklist for evaluating classifiers without hiding the trade-offs that matter.
Published
Writing · Technical notes
Think clearly.
Build with evidence.
Research-informed articles about the decisions behind useful AI systems: what to measure, how to interpret model behaviour, and where confidence should stop.
What explainability can — and cannot — tell you
Feature attribution is a lens on model behaviour, not a shortcut to causality or trust.
Published
These are original educational notes with references to primary documentation. They are not a substitute for a full literature review.