Six axes: drafting an assessment that institutions can use.
Principles become useful when they change a decision. A working assessment should help an institution identify evidence, uncertainty, responsibility and the point at which it must stop.
A standard should improve the conversation before it scores it.
Assessment can become a performance: documents are assembled, boxes are checked and a decision already made is given a responsible-looking frame. A useful standard works earlier. It helps people ask whether a system should be used, for whom, under which conditions and with what route back if the assumptions prove wrong.
The six axes below are presented on the PACT home page as a working draft. Their value is not in their number. It is in the different kinds of evidence they require an organisation to bring into the same room.
The six working axes.
- Human oversight. Who can understand, challenge and interrupt the process, and is that authority protected in practice?
- Transparency. Can people understand that technology influences the decision, what information matters and how to seek an explanation?
- Privacy. Is every collected piece of information necessary, protected and governed throughout its useful life?
- Safety. Which failures are plausible, how are they detected and what happens when the system behaves outside its intended conditions?
- Non-discrimination. Who might be treated differently, which comparisons can reveal that difference and who is missing from the evidence?
- Social impact. How does the system change work, access, wellbeing and power beyond the moment of an individual decision?
An assessment earns trust by making difficult evidence visible, not by making uncertainty disappear.
Evidence has to be attached to ownership.
A question without an owner is easy to answer vaguely. Each axis should identify who supplies the evidence, who challenges it and who accepts the residual risk. Technical teams cannot answer the social-impact questions alone; policy teams cannot substitute for security testing; leadership cannot delegate accountability to a supplier.
The result does not need to be one universal score. A concise record of assumptions, open risks, safeguards and conditions for continued use can support a better decision than a polished number with no route back to the underlying evidence.
What makes the assessment usable?
- Questions are written in language the responsible teams can act on.
- Required evidence is proportionate to the possible harm.
- Unanswered questions remain visible rather than becoming automatic passes.
- The affected people have a route into the assessment.
- Conditions for review, suspension and withdrawal are recorded before deployment.
Responsibility becomes operational when the institution can show not only what it believes, but how those beliefs shape procurement, design, deployment and the decision to stop.