Skip to content
voiage

Qualitative Information Assessment

This experimental workflow helps a review group record which information gaps deserve attention when the evidence does not support a numerical Value of Information calculation. It preserves ordinal priorities, recommendations, dissent, missing fields, source access and a digest-linked, tamper-evident audit history when the digests are externally anchored. It does not turn those judgements into probabilities, utilities, currency, weighted scores or a quantitative VOI estimate.

Run the synthetic reference assessment

Section titled “Run the synthetic reference assessment”

From a source checkout, evaluate the governed fixture and retain the complete versioned result as JSON:

Terminal window
voiage --format json assess-qualitative-information \
  specs/frontier/qualitative-information/v1/fixtures/normative/input.json \
  --output qualitative-information-result.json

Use the text format when a concise, non-colour-dependent rendering is needed:

Terminal window
voiage --format text assess-qualitative-information \
  specs/frontier/qualitative-information/v1/fixtures/normative/input.json

The synthetic example asks whether to run a bounded pilot and whether to refresh a redacted concern register. Two human reviewers agree that the pilot is urgent and should be pursue_now; the register is routine and should be monitor. The renderer preserves the redaction marker without exposing private source content.

What the workflow preserves

Section titled “What the workflow preserves”

Each request declares the decision and alternatives, ordered priority and recommendation classes, reviewers and accountable approvers, source access status, information questions, reviewer judgements, missing fields, redaction state, AI provenance where applicable, and a digest-linked audit sequence.

The evaluator returns complete priority groups and question dispositions. Agreement is reported as consensus; competing classes remain explicit dissent or a complete tie and are not silently resolved. Missing required evidence or unverified AI assistance gives the workflow an explicit unverified state; missing evidence or dissent keeps it incomplete. Approval is reported only when a named accountable human records it after a complete review.

The text alternative includes headings, explicit workflow and approval status, priority groups, unresolved questions, question dispositions and redaction markers. It never relies on colour alone.

Maturity and execution boundary

Section titled “Maturity and execution boundary”

This family is planned for v1.3.0 and remains experimental. Python is the only executable implementation. Rust, R and Julia are unsupported, Mojo is outside the repository boundary, and practitioner, scientific, usability and accessibility review remain required before any stable-promotion claim.

The source fixture is repository-authored synthetic evidence. A completed assessment supports an accountable ordinal workflow only; it is not evidence that a study is cost-effective and must not be described as EVPI, EVPPI, EVSI or another numerical VOI result.