Unceded Intelligence: Delegating Tasks Without Surrendering Authority
Description
Unceded Intelligence: Delegating Tasks Without Surrendering Authority presents Herbert R. Sim’s proposed philosophical framework for benefiting from artificial intelligence while retaining legitimate human authority over its use.
The paper distinguishes capability, operational autonomy and legitimate authority. Its central argument is that superior performance does not independently authorize a system to redefine human purposes, expand its permissions or make withdrawal ineffective.
Through conceptual analysis and hypothetical examples, the paper develops five founding commitments:
-
Human purposes remain open to revision.
-
Delegated powers remain bounded.
-
Consequential decisions remain contestable.
-
Withdrawal remains usable.
-
Responsibility remains attributable.
The discussion connects these commitments to scholarship on meaningful human control, automation, contextual integrity, non-domination and algorithmic auditing. It examines purchasing assistants, AI-mediated representation and essential assistive services, while addressing efficiency, human error, practical dependence and the rights of people beyond the immediate user.
This expanded conceptual working paper includes scholarly references, objections and replies, limitations and a research agenda. It has not been peer reviewed and does not report empirical validation.