Tag
This paper introduces Constructive Alignment, a paradigm that reframes AI alignment as governing the evolution of human preferences over time rather than satisfying static preferences. It proposes a control-theoretic framework to regulate how AI systems influence value trajectories.