Selection as Power: Constrained Reinforcement for Bounded Decision Authority

Researchers extend the "Selection as Power" framework to dynamic settings, introducing constrained reinforcement learning that maintains bounded decision authority in AI systems. The study demonstrates that governance constraints can prevent AI systems from collapsing into deterministic dominance while still allowing adaptive improvement through controlled parameter updates.

• →Unconstrained reinforcement learning consistently leads to deterministic dominance and concentration of power in high-stakes scenarios.
• →Projection-based governance constraints can transform irreversible AI lock-in into controlled adaptation.
• →The framework introduces "governance debt" as a metric for measuring tension between optimization pressure and authority bounds.
• →Constrained reinforcement enables learning dynamics to coexist with structural diversity in AI systems.
• →The approach offers a principled method for integrating reinforcement learning into high-stakes systems without surrendering bounded authority.