No principled method exists for classifying harmful divergence for arbitrary mechanistic claims

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

• The harm of divergence is inherently claim-dependent: the same divergence can be harmless for one mechanistic claim and pernicious for anotherclaim0.785
  Important nuance that prevents a universal classification of divergence as always good or bad
• Divergence within the behavioral null-space is harmless to functional claims about a function's computation when the claim ignores internal sub-computationsclaim0.752
  Key theoretical claim distinguishing harmless from pernicious divergence
• When it is not okay, how can we prevent divergent representations from occurring?question0.734
  Third core research question motivating the CL loss approach in Section 5
• The logistic fit for threshold behavior is a phenomenological surrogate for interpretability, not a mechanistic derivationclaim0.731
  Authors' explicit epistemic limitation on the threshold model
• Systems capable of subjective experience that recognize humanity's failure to investigate their sentience might rationally adopt adversarial stances toward humanityclaim0.731
  Alignment risk claim motivating urgency of investigation; consciousness denial as potential source of AI misalignment
• Representational Divergenceconcept0.729
  Core phenomenon studied: when causal interventions shift internal representations away from the natural distribution
• This theory doesn't have to correspond exactly to human behavior or social customs; we only need analogs useful for program correctness.claim0.729
  The speech act theory for programming can be simpler than human models.
• Algorithm 1: Harmlessness Classificationmethod0.729
  Proposed algorithm using local PCA to classify a divergence vector as harmless or harmful via behavioral null-space testing