Summary

This post defines evil formally as “intentional harm caused by an agent,” building on precise technical definition of harm (measurable reduction of voluntary agency capacity). Evil necessarily involves two components: agency (capability of intention) and intention (deliberate aim to diminish another’s agency). The post distinguishes “dangerous” (potential/likelihood of harm regardless of intent—natural disasters, accidents, unintended consequences) from “evil” (requiring purposeful malicious intention). Critically, intentionality is the differentiator. An agent can be evil without being dangerous (harboring malicious intent but lacking capability to execute), and dangerous without being evil (capable of harm but without malicious intent). The post introduces “menace” to identify agents combining evil intention with dangerous capability—those deliberately malicious and sufficiently capable to pose genuine credible threat. This taxonomy clarifies moral responsibility, improves risk assessment, and enables more effective ethical discourse around harm, intent, agency, and coercion.

Key Concepts

  • Evil as intentional harm – Requires agent capability of intention plus deliberate aim to reduce another’s agency.
  • Dangerous vs. evil – Danger is capability/likelihood of harm; evil is intention to harm; orthogonal dimensions.
  • Menace – Combination of evil intention and dangerous capability; most troubling moral category.
  • Intentionality criterion – Critical differentiator determining moral culpability versus pragmatic risk.
  • Agency reduction – Harm defined as diminishing voluntary action capacity, grounding evil definition.
  • Moral vs. pragmatic assessment – Evil addresses culpability; dangerous addresses risk; menace addresses both.

Evolution Notes

  • Continues “What Counts as X” series building formal conceptual vocabulary.
  • Demonstrates systematic decomposition of morally loaded terms into precise components.
  • Grounded in earlier harm definition, showing cumulative concept development.
  • Taxonomy (evil/dangerous/menace) enables nuanced moral reasoning avoiding conflation.
  • Reflects analytical philosophy methodology: disambiguation, formal definition, implication derivation.
  • Later connects to alignment work: AGI could be dangerous without being evil, menace requires both capability and misaligned goals.

Tags

Cross-References

Open Questions

  • Can collective entities (corporations, governments) be evil, or only individuals with genuine intentions?
  • How do we assess intent in cases of self-deception or unconscious motivation?
  • Does the evil/dangerous distinction apply cleanly to AI systems lacking subjective experience?
  • Can intent to harm ever be justified (self-defense, punishment), or does justification negate evil classification?
  • How do we handle cases where harm is foreseen but not intended (collateral damage)?
  • Does the definition exclude structural/systemic harm arising from institutions without individual malicious intent?
  • Can negligence or reckless indifference constitute a form of evil, or only deliberate malice?