Governing Near Environmental Criticality: The Triple Bind and the Turn to Act-Regarding Restraint

A governance arrangement approaching a critical threshold in the system it governs is in a situation this paper argues is structurally worse than the sum of its difficulties. Three quantities on which such an arrangement depends degrade as the threshold nears, and they are not three independent risks to be managed separately: they are monotone functions of a single approach parameter, so they degrade together and their unfavourable values coincide exactly where governance most needs them. Detection reliability falls, because the statistical signatures on which early warning depends become harder to estimate as the system's recovery slows. The margin within which an intervention is safe shrinks, because the distance to the boundary of the occupied basin is what shrinks as a threshold nears. And estimation of the system's position degrades, in a manner the paper argues is worse than mere imprecision: near-unit-root estimation of persistence is biased downward, and because the recovery rate is a logarithm of persistence, a small downward bias becomes a large overstatement of stability precisely where persistence approaches one. The paper reports a simulation confirming that the same bias which overstates stability by five per cent far from a threshold overstates it nearly threefold near one, so the error runs toward apparent safety and grows as safety diminishes. The governance consequence is the paper's central claim. The standard institutional response to uncertainty, which is to observe more closely, act more cautiously in light of what is observed, and learn from the response, has all three of its legs cut at once by the triple bind, since observing, acting, and learning degrade together and in the same direction. The paper argues that near criticality an arrangement must therefore transfer weight from inputs that describe the system to inputs that describe the proposed act, principally its reversibility, because a property of an act does not degrade as the system approaches a threshold. This is the sense in which the burden criterion of the series' first paper is not merely applicable near criticality but is the only governing input that survives it. The paper surveys the relevant statistical, dynamical, decision-theoretic and institutional literatures, states what each owns, and identifies the residue. It reports no field study.