Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Spider-Man: Across the Spider-Verse illustrating Canon Events Must HappenScene still / Spider-Man: Across the Spider-Verse

Canon Events Must Happen

Miguel treats recurring Spider-Person tragedies as necessary “canon events” and tries to stop Miles from saving his father. The Spider-Society enforces that model as law.

A recurring correlation is treated as certain causation, then converted into coercive policy without a meaningful appeal. The institution becomes more confident than the evidence warrants.

The film leaves the multiverse mechanics unresolved, so Miguel’s model may be partly correct.

AI Risk families

This scenario is an example of the following types of AI risk.
  1. Security
  2. Systemic

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Automation Bias
  2. Calibration
  3. Contestability
  4. Governance Failure
  5. Value Lock-In