The Certainty–Scope Trade-Off in AI: Understanding Floridi's Conjecture

Luciano Floridi (Digital Ethics Center, Yale University / University of Bologna) has proposed a formal conjecture: as an AI system's scope widens, its provable certainty must fall, bounded by a universal constant. Symbolic AI buys certainty by staying narrow; generative AI buys scope at the cost of irreducible error.

This briefing summarizes the conjecture and its grounding in expressiveness-tractability trade-offs, no-free-lunch theorems, formal verification limits, and PAC learning bounds, then translates it into governance terms: policies that assume zero-error AI in open-world settings are mathematically unrealistic, and governance strategy should focus on layered assurance and acceptable risk rather than promising absolute safety.

Attribution: this briefing is a governance-oriented summary of, and commentary on, Luciano Floridi's research and is not a substitute for reading the original paper.