Framework

Six alignment targets

The question it answers

When a team deploys an AI assistant, what should it follow: users' instructions, intentions, revealed behaviour, informed desires, interests or values?

Origin
Gabriel, 2020
Diagram type
Process by phase
Level of analysis
Team
Topics
AI for leaders
Glass sphere with a glowing neural-network core, icon for AI for leaders

Diagram

Six alignment targets

  1. Ladder of targets
  2. Follow instructionsRisk: reading them too literally.
  3. Follow expressed intentions
  4. Follow preferences revealed by behaviourMay be skewed or adaptive.
  5. Follow informed, rational desires
  6. Follow interests and well-being
  7. Follow moral valuesThe most demanding.

Sources

References

Listed A–Z by author surname, Harvard style. A source with a DOI links to https://doi.org/…; one without links to its page or says plainly that it has no DOI.

  1. Gabriel, I. (2020) ‘Artificial intelligence, values, and alignment’, Minds and Machines, 30(3), pp. 411–437.doi:10.1007/s11023-020-09539-2Source page