Research records provisional
Independent research

AI assurance

Assurance research

Did the safeguards
work in practice?

Node & Norm studies the evidence behind human control, decision authority, and correction. Explore what supports a claim that safeguards worked, and what remains unresolved.

01 / Human control

Could a person challenge or stop the decision?

Practical human control

Trust, Autonomy, and Evidence (TAE)

Six conditions examine whether a person could affect an automated decision, from receiving information to making an intervention take effect.

Three selected historical cases and one assessor demonstrate the method. General reliability and transfer to current AI systems remain unestablished.

Publication status: On hold pending arXiv approval

Documented human influence

Human Influence Telemetry (HIT)

A structured assessment asks what the record shows about human judgment and its influence on a decision.

The preserved reviewer agreement concerns one frozen packet and one scoring contract. It does not establish effective oversight.

Research status: Research in progress

02 / Decision evidence

Can the institution reconstruct what it authorized?

Evidence, authority & action

Governed Decision Intelligence (GDI)

A proposed decision record connects evidence, alternatives, uncertainty, human authority, and obligations at the point of action.

The specification and reference implementation are available. Deployment must connect the record to actual controls; field effectiveness remains untested.

Availability: Open source

Decision records in practice

Regulated Gate Decision Support (RGDS)

Structured phase-gate records make claims, supporting evidence, and reasoning available for review.

Illustrative records and validation checks show how the structure works. Institutional benefits and deployment effectiveness remain unmeasured.

Availability: Open source

03 / Incidents & repair

What happened, and what should change?

Emerging research

Incident Reconstruction & Assurance

Source-linked case records separate established facts, allegations, and unknowns to examine where a safeguard may have failed.

These are provisional documentary reconstructions. They are not validated assurance cases or independent investigations.

Six source-linked private draft records; editorial admission pending

Poenitentia Institutionum

Institutional Responsibility & Repair

A proposed institutional framework connects acknowledgment of harm and responsibility with repair for the person and reform of the process.

The work is conceptual and normative. Improved repair or reduced harm has not been established.

Publication status: Editorial review

Follow the claim to the evidence.

These works examine different parts of a decision. Each finding retains its own sources, method, version, and limits. Publication status and open-source availability describe the work; they do not establish that a safeguard is effective.