Skip to content

Incident and Near-Miss Review

A structured retrospective review — instantiates Catastrophic-Risk Bargaining De-escalation

Reconstructs dangerous incidents and the close calls that almost became them to expose the hidden pathways and perverse incentives behind them, then converts each finding into a concrete control or payoff change.

Every standoff that hasn't yet tipped over is generating warnings — the message misread, the order that nearly went out, the moment two units came too close — and most of them are thrown away as "no harm done." Incident and Near-Miss Review treats those close calls as the cheapest and most honest data anyone will ever get about how this dispute could actually kill people, and mines them systematically. Its defining stance is that a near-miss is a free rehearsal of the catastrophe, worth as much scrutiny as a real one, and that the point of the exercise is never to assign blame but to find the latent pathway and the incentive that pushed a person toward risk — then to change the system so the next close call doesn't land. It is retrospective and improvement-driven, which is what separates it from building a shared factual picture in the middle of a live crisis.

Example

During months of tense patrols in contested waters, two coast guards log a near-collision: two vessels closed to within a few boat-lengths before both sheared off. No damage, so the instinct is to move on. A joint review instead reconstructs the minute-by-minute timeline from both bridge recordings and finds three things stacked together — an ambiguous radio call that each side read as the other yielding, a standing "hold your position" order that quietly rewarded the junior captain for not giving way, and a blind arc where neither radar saw the other until late. The findings convert into changes: an unambiguous passing protocol, a revised standing order that credits captains for de-confliction rather than for holding firm, and a sensor-coverage fix. The next tense encounter, months later, resolves in seconds because the incentive to play chicken is gone.

How it works

What distinguishes the review is where it looks and what it produces:

  • Reconstruct from independent sources. Build one defensible timeline from multiple records, not the account of whoever tells it first.
  • Hunt the latent pathway. Ask how close did it come and what almost failed, surfacing the hidden chain that a "nothing happened" verdict conceals.
  • Name the incentive, not the villain. Identify what rewarded the risky choice — a metric, an order, a status fear — because that is what will reproduce it.
  • Close the loop into changes. Every finding becomes an owned control or payoff change, tracked to implementation, and filed so patterns emerge across events.

Tuning parameters

  • Blame-vs-learning stance — how firmly the review protects reporters from punishment. Lean toward learning and you keep seeing near-misses; lean toward blame and reporting dries up, blinding you.
  • Reporting threshold — how minor an event still gets logged. A low threshold catches weak signals early but raises volume; a high one misses the quiet drift toward danger.
  • Joint vs unilateral — reviewing with the other side builds shared understanding and trust but constrains candor; reviewing alone is franker but can't fix cross-party pathways.
  • Analysis depth — proximate cause versus full root-cause. Deeper analysis finds systemic fixes but costs time and can over-reach into speculation.
  • Follow-through tracking — how hard implementation of recommendations is enforced, which decides whether the review changes anything or just documents.

When it helps, and when it misleads

Its strength is that it converts luck into learning: it turns the standoff's free warnings into system changes and, uniquely, exposes the incentive traps that quietly reward brinkmanship. Its failure modes are well known — a blame culture suppresses the very reports the method feeds on, so the organization goes blind precisely where it feels safest, and recommendations that no one is made to implement turn the review into theater.[1] Hindsight bias can make a chaotic event look like obvious negligence, and the exercise is easily run backwards to build a case against a chosen scapegoat. The discipline that guards against this is a just-culture stance that protects reporters, plus hard tracking that every accepted finding actually becomes a change.

How it implements the components

Incident and Near-Miss Review fills only the learning-and-incentive slice of the archetype:

  • near_miss_and_incident_ledger — it builds and mines the accumulating record of dangerous events and close calls, the corpus from which cross-event patterns are read.
  • payoff_and_incentive_repair_plan — its findings about what rewarded the risky move become concrete recommendations to repair those incentives, so the standoff stops paying out for brinkmanship.

It does not build the parties' shared factual picture during a live dispute (joint_fact_finding_record) — that is Joint Fact-Finding Session; it does not display risk in real time (stochastic_catastrophe_risk_state) — that is Residual-Risk Monitoring Dashboard; and it does not model the forward escalation ladder itself (escalation_pathway_and_risk_ladder) — that is Mutual Risk-Reduction Sequence.

Notes

The review depends on a supply of reported near-misses, which is a cultural asset, not a technical one: the mechanism can be perfectly designed and still starve if people fear that reporting a close call will be used against them. Its output is only as good as the follow-through — an unimplemented recommendation is indistinguishable from having never reviewed at all.

References

[1] A just culture distinguishes honest error and system-induced risk (which are learned from, not punished) from reckless or malicious acts (which are not), so that people report close calls instead of hiding them. Near-miss reporting systems in aviation and medicine rest on exactly this bargain; where the bargain breaks, reporting collapses and the warning signal is lost. ↩