Tag: contextual-bandits

Adaptive AI changes the treatment while an experiment is running, which can violate assumptions behind ordinary A/B tests. Good experiments log policy, probability, context, and time.

Contextual bandits can choose among motivational interventions while learning from feedback, but the reward function determines whether adaptation becomes helpful or manipulative.