Educational simulation environment. Do not enter real patient information. AI content supports learning and reflection; it does not replace clinical judgment, local policy, or escalation pathways.

Verification Bench

Verifiable evidence rather than a black-box prediction. Nothing on this platform is shown as a result unless it was actually computed from the synthetic record.

Principles

What must hold before anything is displayed

No fabricated result

Every figure shown on a learning artifact is either genuinely computed from the synthetic record or displayed as an honest “not available” state.

Check: Every numeric field resolves to a computation or an explicit gap marker.

Deterministic before generative

Verdicts come from inspectable rules and computation. Language models may phrase an explanation; they never decide one.

Check: Signal logic is readable in full by the learner.

Adversarial check

A teaching signal is re-run against shuffled labels. If it survives random relabelling, it was never a signal.

Check: Permutation result recorded on the registry entry.

Provenance trail

Each request records a hash of its inputs and outputs so an educator can reconstruct exactly what a learner saw and when.

Check: Append-only audit entries; nothing is deletable.

Privacy-preserving collaboration

Multi-site comparison is designed around sharing derived proofs rather than records. No raw data leaves its site.

Check: No cross-site record transfer in any workflow.

Standards

Sources this platform answers to

WHO
Ethics and governance of artificial intelligence for health (2021)

Ethics, human rights, autonomy, transparency, equity and accountability to health workers.

Open source
ONC
HTI-1 Final Rule

Transparency requirements for predictive decision support in certified health IT: purpose, inputs, fairness, validation, update schedule.

Open source
FDA
Clinical Decision Support Software guidance

Regulatory framing for what is and is not a device function. This platform claims no device status.

Open source
Provenance ledger

Append-only, reconstructable

Illustrative entries from the synthetic environment. Input and output hashes let an educator confirm exactly what a learner saw.

EntryTimeActorActionInput / output hash
AUD-22912026-09-07 14:02 UTCEducator · M. OkaforPublished evidence card revisione3b0c442…98fc1c14 / 9f86d081…a495991b
AUD-22922026-09-07 15:47 UTCLearner · synthetic sessionOpened signal inputs on acute-care case2c26b46b…68e88f7a / fcde2b2e…e5c1b3f8
AUD-22932026-09-08 04:11 UTCGovernance · A. LindqvistEquity review recorded, artifact held at phase 3486ea462…63a1b17f / d4735e3a…88b399d3

Audit entries cannot be edited or deleted from within the platform, including by an administrator.