Skip to content

Human Influence Telemetry v0.2.0 — Public Evidence Pack

Choose a tag to compare

View all tags
@mj3b mj3b released this 16 Jul 21:31
v0.2.0
a325373
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Human Influence Telemetry (HIT) v0.2.0 publishes the first public evidence pack for the standalone repository. It applies the unchanged HIT 0.1.0 specification and assessment schema to three publicly documented institutional decision processes.

Included in this release

  • Dutch childcare-benefits harm-period case study
  • Obermeyer population-health case study, with deploying institutions and the manufacturer assessed separately
  • Cigna PxDx case study, including a designed Command disagreement for later inter-rater testing
  • Four actor-specific machine-readable assessments
  • Public source provenance and explicit evidence boundaries
  • Evidence-gated roadmap through v1.0.0
  • Validation coverage for every public case assessment
  • Updated release, citation, provenance, and DOI documentation

Component versions

  • Repository release: 0.2.0
  • HIT specification: 0.1.0
  • Assessment schema: 0.1.0
  • Dimension catalog: 0.1.0

The component versions remain at 0.1.0 because this release adds evidence artifacts without changing the normative construct model, schema fields, dimension catalog, or scoring semantics.

Demonstrated

This release demonstrates that:

  1. HIT can be applied to heterogeneous public documentary records.
  2. One case can be decomposed into separate institutional actors rather than averaged into one profile.
  3. A near-total IE profile can be represented without treating missing records as demonstrated absence.
  4. Case-derived disagreements can be preserved explicitly for later reliability testing.
  5. Four public case assessments validate against the 0.1.0 schema.
  6. The repository can validate fixtures, negative cases, public assessments, release files, and metadata through one command.

Not demonstrated

This release does not establish:

  • inter-rater reliability
  • prospective institutional effectiveness
  • causal effects on harm, correction, repair, or reform
  • legal liability or compliance
  • standards conformity or certification
  • completeness or truth of institution-controlled records
  • independent institutional adoption
  • superiority over other human-oversight methods

Validation

python -m pip install --requirement requirements-dev.txt
python scripts/validate.py
Assets 2
Loading