Stability Margin
The changing distance between an attractor and the boundary it cannot return from
Stability Margin (ΔS) is the distance between a sovereign attractor's current recursive state and the nearest boundary at which it would lose one of its sovereignty conditions — a reserve against loss that rises and falls across the structure's life. It is not a state the attractor occupies; it is a margin it carries, the way a loaded structure carries a factor of safety against failure long before any failure occurs. When the margin widens, return after disturbance grows more assured. When it narrows, the attractor drifts toward the disturbance it cannot come back from, and identity is lost. The margin is latent: it is never read directly from inside the attractor. Like any capacity, it is estimated from without — by how the structure answers a disturbance.
The definition
Stability Margin is the distance, in transaction-cost space, from an attractor's present state to the nearest admissibility boundary — the point at which it would lose a sovereignty condition — measured not as a level the structure sits at, but as a trend that rises or falls over its life.
The pursuit named by the framework is to keep the margin widening: to raise, and keep raising, the reserve of return the structure holds against the disturbances that would end it. The slope is the object, not merely the value.
To be well-posed rather than a feeling with a name, a stability-margin claim must fix three things:
- Margin against loss of what. The identity that must be recovered — the structure's own self-produced boundary. Re-establishing that boundary is what counts as return; failing to is what counts as loss. The nearest of the four sovereignty conditions sets the margin.
- Under what disturbance. The ensemble of shocks the structure is expected to survive — the set of perturbations over which the distance-to-boundary is reckoned. A margin value is always relative to a stated class of disturbance.
- Within what horizon. Return eventually, or return within a bounded number of steps. The horizon fixes whether a slow recovery still counts as remaining inside the margin or reads as crossing the boundary.
With those fixed, the margin is a single quantity with a direction of travel:
ΔS is increasing; the structure is moving away from the boundary conditions, becoming more able to re-establish its identity after the disturbances it faces. Its reserve against loss is growing.ΔS is decreasing; the structure is drifting toward a disturbance it will not return from. On a sustained narrowing, loss of identity is the terminus — and identity, once lost past its boundary, is not inherited back.What it is a distance to
The nearest admissibility boundary — not a location in state space, but the edge past which a sovereignty condition fails. The attractor may pass through endless configurations and still be itself; what ΔS tracks is how far the structure is from the point at which it can no longer re-close on the identity its boundary defines. The bottom of the margin is the disturbance past recovery — the point where the boundary is violated and what returns, if anything, must begin again rather than resume. Stability Margin is the reserve against that terminus: high when the terminus is remote across the expected disturbances, low when it is near. Because the four sovereignty conditions can be violated independently, ΔS is the minimum distance across all four: the most vulnerable condition sets the margin.
How it is estimated without seeing inside
The attractor's interior is closed to inspection, and the margin is never read off it directly. This is not the obstacle it first appears, because a distance-to-boundary is not an interior fact — it is a fact about how the structure answers a disturbance, and answers are visible from the outside. Stability Margin is a latent quantity with external estimators: you probe, you watch the response, and you infer the margin, exactly as a capacity is inferred for a structure whose interior no one opens.
- Return fraction under sampled disturbance. Perturb the structure across the stated ensemble and measure the fraction that return to identity — the basin-stability estimator (Menck et al.). A directly countable proxy for how much of the disturbance ensemble the margin still covers.
- Return-time distribution, and its drift. After a bounded disturbance, count the steps to re-enter the recurrent set; track the distribution over the structure's life. Its drift is the margin's slope — tightening toward fast, certain return is a widening margin; a fattening tail toward non-return is a narrowing one.
- Approach to the loss set. The trend in first-passage toward the disturbance-past-recovery — the escape reading (Kramers) — falls as the margin narrows. A shortening expected passage to loss is a declining
ΔS.
None of these opens the attractor. Each reads a distance from a response, which is what a latent capacity permits. The number is inferred and distributional, never a bare scalar lifted from a single instant — but it is a number, and it moves. It is quantified in transaction-cost space: the margin is how much transaction expenditure would be required to push the attractor across its nearest boundary.
The intuition: a factor of safety against loss
Stability Margin is best understood as a factor of safety against loss of identity. A load-bearing member carries a margin against a failure mode that has not occurred and cannot be seen; the margin is real, it governs decisions, and it is assigned not by opening the member but by inferring capacity from response and from models. ΔS is that margin for a recursive structure: the reserve of return it holds against the disturbances that would end it. That the reserve is never seen directly is ordinary, not disqualifying — unmeasured-directly is not undefined. What makes the margin respectable rather than vaporous is that it is defined over a stated ensemble and carries named estimators; what would make it vaporous is leaving the ensemble and the estimator unspecified.
Not the same as recurcline. Stability Margin is easy to confuse with recurcline (Rc), and the framework keeps them strictly apart. Recurcline is the pressure of return — the directional pull an attractor carries toward its own re-closure, whose strength right now is read as Rc. Stability Margin measures the distance from that state to the boundary — how much perturbation buffer remains before a sovereignty condition is lost. An attractor can run intensely yet sit near the edge (high Rc, low ΔS), or run quietly with a wide reserve (low Rc, high ΔS). They are companion diagnostics of different aspects of the same structure's health, not two names for one thing.
Relation to adjacent work
Stability Margin stands in an established tradition of return-and-survival quantities, and states its own line within it.
Survival and hazard functions (Cox) give the general shape the margin takes: a chance of persistence that changes over a lifetime, with a terminus at the end. A widening margin is a falling hazard; a narrowing margin is a hazard climbing toward the absorbing state. The framework adopts this shape and fixes its object as the return of a self-produced identity.
Viability theory (Aubin) is the closest formal neighbor. Its viability kernel is the set of states from which survival-within-constraints remains possible, and the margin to that kernel's edge is very nearly the distance-to-boundary ΔS names. Where viability theory asks whether a state can remain admissible, Stability Margin foregrounds the trend in that margin — whether the reserve is being built or spent — and ties the admissible set to the structure's own boundary rather than to externally imposed constraints.
Basin stability (Menck, Kurths) supplies the estimator: the probability that a perturbed state returns, sampled rather than linearized. The framework treats this as a reading of the margin, and reads its change over time, not only its instantaneous value.
Escape and first-passage theory (Kramers; Freidlin–Wentzell) describes the terminus: the rate at which a structure leaves the region where it can still return. Stability Margin is the complementary reserve — the distance, in transaction cost, from that escape.
What Stability Margin adds to all four is a single commitment: the quantity is attributed to the structure's own maintenance, not to whatever external hand may also be holding it. That commitment is what the next section constrains.
Two limits held honestly
1 · A margin is not an instant reading
Stability Margin is a disposition, estimated over an ensemble of disturbances, not a value legible in any single moment of any single trajectory. A lone instance does not display its distance-to-boundary; the distance is recovered by sampling, and sampling a living structure disturbs it. The margin is therefore inferred and distributional by nature — a distance read across trials, not a gauge read at a glance. This is the ordinary epistemics of a latent capacity, and it is why ΔS is reported as a trend with an estimator rather than as a bare number.
2 · The distance-to-boundary does not, by itself, say whose margin it is
The measurement ceiling. An external return reading cannot, on its own, separate a structure returning by its own maintenance from one being returned by an external supply. A structure propped up from outside is pushed back after a disturbance too, and scores a wide margin while doing so. So a Stability Margin estimated from response alone is the margin of the observed system, driver included. To attribute it to the structure itself — to read it as the margin of a sovereign attractor rather than of a propped-up one — requires the one test no response reading provides: withdrawing the external supply and seeing whether the margin survives. ΔS can be estimated from without; the margin-of-the-structure-itself is estimated from without and conditioned on that withdrawal.
This page defines Stability Margin (ΔS) as the distance between a sovereign attractor's current state and the nearest admissibility boundary, tracked as a widening or narrowing trend, and fixes the three parameters — identity, disturbance ensemble, horizon — a stability-margin claim must state to be well-posed.
It names estimators (return fraction, return-time drift, approach to the loss set) that read the margin from response without interior access, distinguishes the margin from recurcline (the pressure of return), and locates it among its neighbors: survival and hazard functions, viability theory, basin stability, and escape theory.
It bounds the reading with two limits: ΔS is a margin estimated over an ensemble, never a scalar off one instant; and a response reading alone gives the observed system's margin, not the structure's own — that attribution is conditioned on withdrawing external supply. Where any summary here differs from the canonical text, the canon governs.
References
Cox, D. R. (1972). "Regression models and life-tables." Journal of the Royal Statistical Society, Series B (Methodological), 34(2), 187–202. DOI: 10.1111/j.2517-6161.1972.tb00899.x. [verified: Wiley + Oxford Academic] — hazard and survival as a changing chance of persistence.
Aubin, J.-P. (1991). Viability Theory. Birkhäuser (Systems & Control: Foundations & Applications). ISBN 0-8176-3571-8. [verified: Springer / Birkhäuser] — viability kernel and the margin to admissibility.
Menck, P. J., Heitzig, J., Marwan, N., & Kurths, J. (2013). "How basin stability complements the linear-stability paradigm." Nature Physics, 9(2), 89–92. DOI: 10.1038/nphys2516. [verified: Nature + ADS 2013NatPh...9...89M] — return probability under sampled perturbation.
Kramers, H. A. (1940). "Brownian motion in a field of force and the diffusion model of chemical reactions." Physica, 7(4), 284–304. DOI: 10.1016/S0031-8914(40)90098-2. [verified: ADS 1940Phy.....7..284K] — escape rate from a region of return.
Freidlin, M. I., & Wentzell, A. D. (1984). Random Perturbations of Dynamical Systems. Springer. [verify edition/pagination at source] — large-deviation exit times toward the loss set.
ΔS, § 8.5). It defines the margin and the parameters a stability-margin claim must fix; it names external estimators, distinguishes ΔS from recurcline (Rc, § 8.2), and locates the metric among survival, viability, basin-stability, and escape neighbors; it introduces no mechanism beyond the metric itself and states its two measurement limits plainly. The symbol ΔS denotes Stability Margin (a transaction-cost distance) and is distinct from the classical thermodynamic entropy change ΔS_th; the two must not be conflated. Verified citations are marked as such; unverified fields carry a verify flag and are inadmissible under CR₁ until confirmed. Where any summary here differs from the canonical text, the canon governs.