FlightReady
A go/no-go risk model for general-aviation flights that shows its work — every score decomposes into factors, uncertainty, and tail risk.
A research prototype for standardized aviation go/no-go decisions. It takes the inputs a preflight risk assessment already uses — weather, pilot currency, route and terrain, aircraft, time of day — and turns them into a transparent risk score with per-factor contributions, an uncertainty band, and a tail-risk (CVaR) view of the worst outcomes, alongside a structured briefing. Designed around human-in-the-loop review, not automation of the decision.
The problem
General-aviation pilots already do preflight risk assessment. Most fill out some version of a Flight Risk Assessment Tool — score weather, pilot currency, route and terrain, the aircraft, time of day, land in a green / yellow / red band. It's a genuinely good habit: it replaces a vague feeling with a number and forces you to look at each factor on purpose.
But the number it produces is an average, and the standard 5×5 matrix is coarse — it collapses a whole distribution to a single point, and it jumps between bands for reasons that have more to do with rounding than with the sky. A go/no-go call is not a bet on the average day.
The thesis
A go/no-go call is a bet on the worst plausible day, not the average one. The tool should show the tail.
Four models, one set of inputs
Rather than pick a single "correct" risk formula, FlightReady runs four — because the point is to build intuition about what each one conceals. You can operate all of them on the same flight in the aviation risk explorer:
- Simple matrix — the FRAT baseline. Chunky and discontinuous; nudge one input and the score can jump a band or not move at all.
- Weighted — smooth, and it decomposes: you see which factor is actually driving the risk, so you can tell whether it's broad or concentrated in one thing you might mitigate.
- Probabilistic — a seeded Monte-Carlo run varies the uncertain factors and produces a p5–p95 band instead of a single point. Same mean, honest spread.
- Tail-risk (CVaR) — the one that matters. It reports the average of the worst outcomes, not the average outcome.
Why CVaR
Quantitative finance ran into this exact problem and mostly moved past it — from Value at Risk (a threshold that says nothing about how bad the tail beyond it is) to Conditional Value at Risk, the mean of the worst outcomes. Bring that lens to a flight and a volatile-weather day stops hiding behind a benign average: its mean might match a steady flight, but its CVaR is much higher, because the worst tenth of its simulated days are genuinely dangerous. The number finally reflects the question the pilot was actually asking. The full argument is in tail risk in aviation decisions.
The through-line to the rest of my work
This isn't a detour from the legal work — it's the same idea. A risk model that surfaces the tail and abstains at the edges is exactly what Littman does when it routes a high-severity-if-wrong decision to a human instead of guessing. High-stakes systems should expose their uncertainty and keep a person in the loop, whether the stakes are a contract or a Cessna.
Status & limits
Prototype. The weights and distributions are illustrative, not calibrated against incident data — honest calibration would need real data and far more care. What exists today is the interface and the risk math, built to make a specific argument you can hold in your hands.