Memory Refresh Through Succession
Venice's roughly thousand-year persistence came from converting a rare, long-horizon hazard into frequent, short-horizon surrogate events — the doge's promissione ducale renegotiated at every succession, and lot-and-vote elections designed to make office capture impractical.
What decision changes?
Test whether a proposed refresh mechanism (drill, audit, sunset clause, succession) actually exercises the real handle under conditions the target cannot fully pre-optimize and can be observed to fail — a permanent zero failure rate is evidence of decoupling, not health.
Human collective memory has a time constant, and it is short: roughly a working lifetime. A hazard that recurs less often than that becomes invisible — not because anyone decides it is unimportant, but because the evidence confirming the danger dries up, while the evidence that “we relaxed the rule and nothing happened yet” keeps accumulating. Every safeguard against a rare disaster is therefore in a slow-motion race against its own obsolescence in the minds of the people maintaining it.
The Republic of Venice is history’s most instructive answer. It lasted roughly a thousand years and fell to Napoleon’s army, not to internal capture — a persistence record almost no other republic approaches. Its trick was never unusually long memory. It was converting the rare hazard (a doge seizing permanent personal power) into frequent, short-horizon events that individual careers could experience. The doge’s promissione ducale — the contract constraining the office — was renegotiated at every succession rather than inherited passively, with each new version incorporating restrictions drawn from the previous reign’s abuses. The constraint was thus re-derived, every generation, by people with a live stake in it. After the 1310 Tiepolo conspiracy, Venice added an election procedure of alternating lotteries and votes so elaborate that buying or packing the outcome became practically infeasible. Successions, drills, sunset clauses requiring active reauthorization, the Catholic Church’s advocatus diaboli (the institutionalized skeptic in canonization proceedings) — all are versions of the same move: manufacture small surrogate events that keep a rare hazard inside the horizon of living attention.
The Church case even supplies a natural experiment: when the devil’s-advocate role was abolished in 1983, canonization rates rose roughly tenfold. Remove the ritualized corrector, and the throughput it was checking floods through.
But this mechanism has a failure mode that is easy to manufacture by accident, and it matters directly for AI oversight. A refresh ritual keeps genuine contact with the hazard only if it exercises the real handle, under conditions the audited party cannot fully anticipate, and can visibly fail with consequences. An announced fire drill everyone prepares for, an audit conducted on the auditee’s schedule with the auditee’s chosen evidence, a drill failure rate of exactly zero over many years — these pass the ritual’s surface while the exercised handle has quietly decoupled from the real one. For AI evaluation regimes, the test is the same: an eval the developer schedules, scopes, and never fails is not evidence of safety; it is evidence the ritual has come loose from the hazard.
One of eleven historical case studies in Institutional Genesis, Memory, and Decay — see the overview for the full life-cycle map, or read the complete appendix.
What would count as evidence?
The Catholic Church's advocatus diaboli role in canonization, abolished in 1983, was followed by a roughly tenfold increase in canonization rates — a natural experiment in what happens when a ritualized corrector is removed rather than merely weakened.