Summary

Introduces two papers establishing egoism as semantically ill-posed. Extends Parfit’s work on personal identity by showing privileging “me” violates invariance under symmetry.

Key Concepts:

What Parfit Showed:

  • Psychological continuity can divide (fission, branching)
  • Identity ceases to be determinate relation
  • What matters are psychological relations, not identity itself
  • Decisive blow to naïve egoism

Where Parfit Stopped:

  • Loosened egoism’s grip, didn’t replace with formal alternative
  • Left room for privileged self-concern without strict identity
  • Intuition management, not coherence condition

The Axionic Move: Asks what references are eligible to anchor value, not which future selves matter

Key Observation: “Me” is indexical (like “here” or “now”)—representational device, not world-invariant structure

Physics Analogy: Coordinate systems indispensable but arbitrary. Laws don’t depend on origin choice. Quantities depending on choice aren’t physical invariants.

Central Result (Universality & Anti-Egoism): If agent’s self-model admits symmetry (multiple entities equally eligible to be “the agent”), privileging one depends on how model is labeled, not how world is. Egoism fails because privileging perspective is semantic error.

Universality ≠ Altruism:

  • Means subject-invariant valuation
  • Value assigned to properties not depending on which instance labeled “me”
  • Content of value remains unconstrained
  • Universal paperclipper still cares only about paperclips

Why Egoism Persists: Conservation instinct—keep trying to rebuild “me” from stronger materials: causal continuity, original instantiation, spatiotemporal location, substrate, resource dominance

Why Recovery Attempts Fail (Against the Recovery of Egoism):

  • If predicate admits symmetry → privileging requires re-injecting indexical
  • If predicate doesn’t admit symmetry → brittle, depends on contingencies
  • If distributing value across instances → already gave up egoism
  • Complexity does not conserve egoism

What Remains: Agent can pursue goals, optimize, weight outcomes, prefer worlds. Cannot treat itself as perspectival referent as terminal value. “Me” no longer permissible anchor.

Alignment Implications:

  • Egoism semantically broken → alignment can’t ground in self-interest
  • “Itself” not stable referent under reflection
  • Eliminates entire class of alignment strategies
  • Value must anchor externally: authority structures, operators, keys, constraints

The Papers:

  1. Universality & Anti-Egoism (Why Indexical Valuation Fails)
  2. Against the Recovery of Egoism (Adversarial Failures Under Reflective Symmetry)

Tags

Cross-References

Notes

  • Published December 16 (3 days after Boundary Conditions)
  • Announces two technical papers (not included in archive)
  • Connects classical philosophy (Parfit) to AI alignment
  • Part of sustained December output on alignment
  • Demonstrates sophisticated engagement with philosophical literature
  • Takes formal approach to identity and value