Avatar Intelligence: amplifying human agency with persistent personal agents

The paper proposes the Avatar Intelligence paradigm: amplifying human agency with persistent personal agents that represent the user's goals and constraints across situations and over time, and measuring progress by less routine supervision. The authors present a maturity roadmap from human-led execution to authorized autonomous operation within a domain, and propose an Agentic Oversight Index (AOI) that weighs delivered value, operating cost and human supervision burden. The research agenda focuses on identity-continuous personalization, Rehearsal Engineering, and safety mechanisms combining technical control with accountability and remedy.