Linch introduces 'Frame Errors' as third kind of reasoning mistake

Linch's new LessWrong post 'Frame Error' introduces a third category of reasoning mistakes distinct from logical fallacies and empirical errors. These 'frame errors' occur when the entire framework for reasoning about a problem is structurally inadequate—even if every fact is right and every inference valid. The post provides five common classes with worked examples, starting from well-known concepts like Pauli's 'not even wrong' and progressing to novel instances where Linch claims to be the first to explicitly identify the error. Conceptual tools are offered to help readers model good thinking and detect when they or others are committing frame errors. The framework is particularly relevant for fields like AI safety, where subtle misframings can lead to catastrophic mistakes.

While the post appears to have been inadvertently published as a draft (as noted in comments by Rachel Shu and Linch), the core idea is gaining traction in rationality circles. Linch differentiates his concept from 'not even wrong' by emphasizing that frame errors are a distinct category of mistake that can be identified and corrected. The post concludes with advice on avoiding meta-errors—errors about which framework to apply. For AI researchers and developers, this framework offers a new lens to scrutinize problem formulations, potentially uncovering blind spots in current reasoning about systems like large language models or value alignment.