How can we produce a principled method for classifying harmful divergence for any mechanistic claim?

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 anotherclaim0.818
  Important nuance that prevents a universal classification of divergence as always good or bad
• Representational Divergenceconcept0.760
  Core phenomenon studied: when causal interventions shift internal representations away from the natural distribution
• Harmless Divergenceconcept0.753
  Divergences that occur in the behavioral null-space and do not affect functional claims about the model
• When, and to what extent, is it okay for divergences to occur?question0.753
  Second core research question motivating the theoretical analysis in Section 4
• When it is not okay, how can we prevent divergent representations from occurring?question0.748
  Third core research question motivating the CL loss approach in Section 5
• Divergence within the behavioral null-space is harmless to functional claims about a function's computation when the claim ignores internal sub-computationsclaim0.746
  Key theoretical claim distinguishing harmless from pernicious divergence
• The logistic fit for threshold behavior is a phenomenological surrogate for interpretability, not a mechanistic derivationclaim0.742
  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 humanityclaim0.740
  Alignment risk claim motivating urgency of investigation; consciousness denial as potential source of AI misalignment