Provenance, privacy, and limitations
No account, analytics, network requests, or model calls. Run data are stored in this browser when permitted. Export a copy before changing browsers or clearing storage. Do not enter identifiable student information.
Author controls are not access-controlled. This is a teaching and discussion tool, not a secure assessment system or a validated prediction model. Written reasoning is recorded but is not machine-scored.
Model rationale and evidence boundary
Feasibility, authority, access and observability are explicit constraints. Declining, pausing and changing leverage are decisions, not random failures. Outcome ranges are authored scenario assumptions, not empirically calibrated probabilities. The adverse/central/favorable choices are sensitivity tests; no probability is assigned to those futures. Learning impacts remain synthetic, and process metrics cannot establish causal benefit.
Methodological references: RAND: robust decisions under deep uncertainty; HM Treasury: handling complexity in evaluation. These support the approach, not the numerical settings.