Ontological Governance Observatory
Load results from Narrative Drift and SelfTrace and read them as a Second-Order Assessment: what the instruments recorded about how people know, how they present themselves, and how they act, set beside the provisions that address each question.
01
Run an assessment
The report asks one question per layer, lists the numbers each instrument produced, and gives a plain reading that follows from those numbers alone. Everything happens in your browser. No result leaves this page unless you export it.
Load results
Drop one or more result files here, or choose them from your device. Narrative Drift and SelfTrace both offer "Download result (JSON)" at the end of a run. Files are read in this browser and are not uploaded anywhere.
The examples are reference runs: Narrative Drift Script A and Script B (verifier replays of fixed scripts) and the SelfTrace demo profile. See "Verify" below to reproduce them.
02
Verify
The example results are reference runs: exports from particular builds of each instrument. You can reproduce them and check the report against a committed copy.
- Narrative Drift, Script A (verifier replay, sequence 00000-000000000). Open the live instrument. Open the link, let the replay finish, then choose “Download result (JSON)”. The reference file (narrative-drift-script-a.json) was exported from build 51d91e3.
- Narrative Drift, Script B (verifier replay, sequence 00000-222222222). Open the live instrument. Open the link, let the replay finish, then choose “Download result (JSON)”. The reference file (narrative-drift-script-b.json) was exported from build 51d91e3.
- SelfTrace, Demo profile “The Emerging Public Intellectual”. Open the live instrument. Choose the demo profile “The Emerging Public Intellectual”, open /results, then choose “Download result (JSON)”. The reference file (selftrace-demo-intellectual.json) was exported from build 09959e8.
- Load your downloaded files above. The report's Sources table shows the instrument build and the SHA-256 of each file. If the build matches the reference file's, the numbers should match; a later build may differ because the instrument changed.
- Compare with the expected report for the three reference runs. In the repository, npm test rebuilds that report from the reference files and fails if anything except generated_at differs.
03
Why this project exists
Governance frameworks assume a stable human subject, one who can consent, deliberate, and hold systems accountable. AI systems are quietly changing the categories through which that subject understands itself.
The Observatory is a small step toward making that change visible. It reads result files from single runs of Narrative Drift and SelfTrace and sets out what they recorded, one question per layer. It does not observe people over time or across a population.
Without sustained observation of these shifts, governance stays first-order. It regulates system outputs while the subject who is supposed to do the governing is reshaped underneath.
A second-order governance watches not only what systems do, but what they do to the conditions of human agency itself.
04
What it reads
Each layer of the framework is paired with one question and the instrument that can speak to it. Every indicator is labelled with the concept it stands in for: the instruments measure proxies, not the layers themselves.
Epistemic: delegation
Narrative Drift records how often routine choices were handed to the system, and where the autonomy axis ended.
Ontological: rewarded self-presentation
SelfTrace compares topics the environment rewarded and that grew with topics the person kept without reward, and tracks topic range over time.
Political: deliberation in civic acts
Narrative Drift records whether choices in the civic encounters shortened or lengthened deliberation, and where the tempo axis ended.
Data handling check (not a layer)
A check on the assessment itself, outside the three layers. Results carry option indices, counts and topic labels; the Observatory refuses files that declare personal content, runs a heuristic scan for free text and processes everything on the device.
05
What the report is not
It is not a legal or clinical assessment. Provisions are named because they address the same question, not to suggest that any product complies with them or breaches them.
The results reflect what happened inside the instruments, which use simulated encounters and, in the examples, a fictional profile. They do not describe how a person behaves elsewhere. A reading says what the numbers show and stops there.
The Suite