A motivation selection method is method that attempts to prevent undesirable outcomes from artificial intelligence by influencing what the AI wants to do. Motivation selection methods encompass direct specification, domesticity, indirect normativity, and augmentation. Motivation selection methods may be contrasted with capability control methods, which attempt instead to restrict what the AI can do.
Bostrom, Nick (2014) Superintelligence: Paths, Dangers, Strategies, Oxford: Oxford University Press, pp. 138-143.