Written enquiry →

THE DECISION FIELD

Selected work from Decision Engineering.

Flagships, research notes and field observations on consequential AI systems, authority, evidence and decision systems under changing conditions.

Selected public work · Curated archive · Research-led practice

THE DECISION FLAGSHIP

Long-form work for problems that deserve more than a short answer.

ISSUE 09 · 6 OCTOBER 2026

THE REVIEW HAPPENED ON TIME. THE DECISION WAS ALREADY LATE.

Review timeliness is not the same as correction timeliness.

THE PUBLIC QUESTION

When does governance remain procedurally correct but become operationally too late?

KEEPER DISTINCTION

Review timeliness is not correction timeliness.

ISSUE 08 · 29 SEPTEMBER 2026

NOTHING CHANGED IN THE SYSTEM. THE SYSTEM WAS NO LONGER AUTHORISED.

ISSUE 07 · 22 SEPTEMBER 2026

THE CHANGE WAS REAL. THE DECISION STAYED CLOSED.

Why detecting change is not the same as reopening the decision that still governs what the system is allowed to do.

THE PUBLIC QUESTION

When does a detected change become significant enough to reopen the decision that still governs execution?

KEEPER DISTINCTION

Detection is not reopening.

ISSUE 06 · 15 SEPTEMBER 2026

THE REVIEWERS DISAGREED. THAT WAS THE RESULT.

When disagreement stops being noise and starts testing the instrument.

Same records. Same questions. Same scales. Different judgments.

What became unstable — the stimulus, the scale, the framing, or the interpretation?

Independent review should be tested, not presumed.

ISSUE 05 · 8 SEPTEMBER 2026

THE GATE WORKED. THE BASIS BEHIND IT DID NOT.

ISSUE 04 · 1 SEPTEMBER 2026

THE WARNING SURVIVED. ITS EFFECT ON EXECUTION DID NOT.

RESEARCH NOTES

Questions that need a closer evidentiary look.

Shorter research-led notes on evidence, judgment, authority and the boundaries of what a result actually supports.

Explore deeper research →

CURRENT DISTINCTIONS

  • Independent review is not a headcount.
  • Current state ≠ state lineage.
  • Valid credential ≠ current authority to execute.
  • Epistemic standing ≠ decision standing.
  • Aggregate performance evidence ≠ authority across all operating conditions.
  • A precise measurement of the wrong evaluation object remains the wrong measurement problem.

FIELD NOTES

Operational distinctions that become visible in practice.

Compact observations from public field work, organisational discussion and recurring decision-system failure modes.

  • Tool adoption ≠ workflow migration.
  • Readable ≠ executable.
  • Valid credential ≠ current authority to execute.
  • Current state ≠ state lineage.
  • Source of truth → source of intervention.
  • Locally acceptable permissions can compose into globally unacceptable authority.
  • Valid claim ≠ authorised actor ≠ admissible consequence.
  • Bound authority first. Expand only when evidence supports expansion.

Public field work is not the research archive.

Decision Engineering is the applied organisational layer. Decision Epistemology is the emerging research programme concerned with the epistemic integrity of decision systems over time.

Explore Decision Epistemology ↗

Does a current decision still fit current conditions?

If the question has operational consequence, examine the authority before the consequence becomes the first visible signal that something changed.

Explore the Review →Send a written enquiry →