The classifier · a walk through the SORT framework

Run the classification on your own numbers.

The walkthrough classifies one case. Give this tool your own point estimates for harm and exposure across two periods, plus how uncertain each is, and it runs the same probabilistic classifier — the whole thing lives on this one screen.

Hover a quadrant to learn more
Trajectory classification quadrant chart with the Mitigating quadrant highlighted.A two-by-two grid. The horizontal axis is the exposure trend (E), decreasing on the left and increasing on the right. The vertical axis is the harm-per-exposure trend (Ĥ), decreasing at the bottom and increasing at the top. The four quadrants are Concentrating (top-left), Escalating (top-right), Receding (bottom-left), and Mitigating (bottom-right).ConcentratingEscalatingRecedingMitigating9.8%58.5%exposure growth →decreasingincreasingharm trend ↑
31.7%Unclassifiable — evidence too uncertain to place
Most likely outcomeexposure ×3.00 · harm/exposure ×0.56

Mitigating — 58.5%

More people are exposed, but harm per unit exposure is decreasing — existing safeguards appear to be working. Continued monitoring warranted; a failure of current controls could shift the trajectory to escalating.

Your estimates

Pick two comparison periods for one monitoring question and give your best point estimate for each quantity. Counts can be raw numbers — only the ratios between periods matter.

Harm (H)

Harmful events matching the question — the numerator. Estimate the level in each period from the strongest sources you have.

Exposure (E)

Opportunities for that harm — the denominator. Uses, interactions, or population that match the monitoring question in each period.

Uncertainty

How much you trust the estimates. Bigger factors and a wider band mean more draws land Unclassifiable rather than forcing a false verdict.