Anthropic Interpretability Team: 171 emotion vectors causally influence behavior; performing vs having functional emotion representation are measurably different

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Anthropic is extremely conservative in writing up interpretability results due to Overton window concerns.claim0.779
  Antra's explanation for why even stronger evidence may exist but remains unpublished.
• Emotion refers to a state concept, so stateful representations in general may be more persistent across tokens.claim0.761
  Interpretive hypothesis offered to explain why emotion features are more persistent
• Anthropic's model-welfare program signals frontier labs taking "what's it like to be a model" seriously, creating space for external measurement.claim0.759
• Representation geometry causally shapes behavior; activation and behavior manifolds are approximately isometric.claim0.759
• To what extent is emotion feature persistence driven by genuine internal emotional state versus autoregressive conversational context dynamics?question0.759
  Core open question the paper raises but does not fully resolve
• Steering vectors discover effective triggers such as 'However' and 'Otherwise', consistent with prior reported reflection datasetsfinding0.759
  Validates that steering vectors capture reflection semantics by finding tokens reported in related work.
• Causal abstraction is not enough for mechanistic interpretability because it becomes vacuous without assumptions about how models encode informationclaim0.758
  Central thesis of the paper
• Text-based and self-steered emotionality ratings are only weakly correlated (ρ = +0.051, n.s.), suggesting they measure different aspects of feature emotionality.claim0.751
  Finding that the two evaluation modalities frequently diverge in their interpretation of the same SAE feature