Diagnostics

For operators, governance teams, and auditors. Three of the four tools score what you enter about a workflow. The record conformance checker grades the records a system exports.

Which tool for your system

Pick a tool by the question you have

  • Scores your answers · 10–15 minutes

    Is workload concentrating on a few people or steps?

    Rate seven sources of workload with the team and see where it concentrates.

    Run the workload modeler

  • Scores your answers · 20–30 minutes

    Is the system still justified in making this decision?

    Take one workflow through six questions and list the permissions nobody on your team can justify.

    Run the delegation audit

  • Checks exported records · 5 minutes

    Do your decision records meet the level you claim?

    Check the records a system exports against the published schema, and recompute the hash chain that links them.

    Open the record conformance checker

  • Scores your answers · 10–12 minutes

    Can the institution correct what it can decide?

    Answer five questions about how the system has grown and how the institution fixes its mistakes.

    Run the corrective capacity self-assessment

If a system decided about you

Short checks: can you challenge it, can the appeal work, who can reverse it

Checks for the person on the receiving end of a decision, to run before escalating.

Ethotechnics is an open standards body, not a legal clinic. It does not take cases. Where you are and what was decided may already entitle you to the reasons, a review by a person, or a decision by a deadline; the standards on this site are drafts and have no legal force. What you can ask for now, and questions to send about the decision.

Can this decision be challenged?

Seven yes/no checks on whether you can challenge the decision.

Check the decision →

Can this appeal change the outcome?

Checks whether an appeal can change the outcome or only records it.

Check the appeal →
How the tools map to standards

  • Workload modeler

    Symptom
    Staff absorb the system's errors where nobody counts them: burnout, manual overrides, hours spent on workarounds.
    Output
    Burden score and where the load concentrates

  • Delegation audit

    Symptom
    Nobody clearly owns an outage, and nobody has rehearsed the fallback.
    Output
    Permissions nobody can justify

  • Corrective capacity self-assessment

    Symptom
    The institution can fix fewer of its mistakes each year, and the system gets harder to back out of.
    Output
    Self-reported concern score

  • Record conformance checker

    Symptom
    The records do not show who authorized each action, or whether a person can challenge it.
    Output
    Conformance level the records earn

What every diagnostic output includes

Every output carries the same baseline and passes the same four checks against being used to cut access or raise burdens.

Each tool answers one question

Each diagnostic answers one question about one workflow, and its method card states what it does not measure.

A readout you can file

Every tool returns a readout you can link, copy, or export and keep with the decision it informed.

Named mechanisms

Recommendations use the mechanism names from the library, so the team that acts on a finding and the team that reviews it use the same terms.

Deliverable baseline

  • Results you can copy, link to, or export, depending on the tool.
  • References to the mechanisms each mitigation comes from.
  • Facilitated follow-up when the question or risk stays unclear.

Before publishing an output about appeals or reversals, run the challenge checklist against the appeal path it describes.

Checks against misuse

Four constraints keep a diagnostic output from being used to deny care, reduce access, or raise burdens.

  • AW-01: Burden metrics include a remediation plan and staffing allocation.
  • AW-02: No appeal path is removed or weakened without a published replacement and a review window.
  • AW-03: A cut to access or eligibility always comes with a receipt and a deadline for remedy.
  • AW-04: High-burden flows include assisted channels or human escalation.

Reference: Anti-weaponization constraints and the evidence requirements.

You bring one scenario and the decision it feeds

A facilitated session covers one scenario. It ends when the decision it was framed around has an answer or a named blocker.

  1. Frame the question

    We agree on what a usable answer looks like and which decision it feeds.

  2. Run the tool together

    You answer the prompts and supply the inputs. We record each gap and who owns it.

  3. Leave with a next step

    You leave with a linkable readout, the mechanisms it cites, and one named next step.

Every diagnostic runs without a prior relationship, and the Institute publishes each one under CC BY-SA 4.0.

Paid work

Commissioned evaluation of healthcare AI

Kanav Jain, who wrote this framework, does commissioned safety evaluation for healthcare AI: a one-hour safeguards review, a two-week readiness sprint on one workflow, an investor diligence memo in five business days, or a two-to-three-week evaluation against FDA and EU AI Act expectations.

Engagements on kanav.net Submit community feedback

Worksheet

An instrument that takes numbers rather than a workflow

It asks for figures you already have and returns a ceiling you can argue with.

Cite this page
More formats

Retired tools and historical source →