Human-position analysis, diagnostics, and a bounded Governance Finding.
AEGIS / ANALYTICAL FRAMEWORK
Governance must be possible before it can be claimed.
ANALYTICAL OBJECT
One bounded claim, tested against operative capability.
“Human in the loop” compresses several different roles into one phrase. A person can authorize a deployment, monitor a system, receive an alert, press a stop control, or respond to an event. Each describes a different relationship to the execution under review.
AEGIS tests the defined claim against architecture, timing, operator conditions, intervention mechanics, and trace. Interface access matters when it affords legible, actionable control before material commitment. Post-event observation supports a different, narrower claim.
RECONSTRUCTION → ATTRIBUTION
AEGIS evaluates a reconstruction whose evidentiary limits remain visible.
VIRGIL qualifies the trace and reconstructs the claim-relevant execution. AEGIS receives that bounded factual predicate, including branches and unresolved relations, then evaluates the governance claim.
VIRGILWhat can the trace support?AEGISWhat governance attribution does the established execution support?
Trace qualification and guided execution reconstruction.
01 / HUMAN POSITION
Presence describes location. Position describes capability.
AEGIS distinguishes the role assigned to a human from the role the architecture, timing, information, and controls made possible during the execution under review.
HUMAN POSITION / HP-01
Pre-Execution Governor
Receives legible information and retains authority and practical means to prevent, redirect, or authorize the action under review.
- Operational interval
- Before material commitment
- Governance-claim boundary
- May support a governance claim when the timing, control path, and available evidence demonstrate that the intervention opportunity was real.
02 / EXECUTIONAL TEST
Timing changes the human position.
The decisive question is whether meaningful alteration remained available during the interval assigned to human governance.
SEQUENCE / APotentially governable interval
- 01System selects action
- 02Legible signal becomes available
- 03Intervention remains actionable
- 04Execution commits
The signal arrives while a meaningful intervention remains available over the execution under review.
SEQUENCE / BPost-commitment intervention
- 01System selects action
- 02Execution materially commits
- 03Alert becomes available
- 04STOP command is recorded
The response may support recovery, mitigation, or later control. It does not establish governance over the action already committed.
03 / ANALYTICAL CONCEPTS
A controlled vocabulary for a precise claim.
01Material Commitment+
The point after which the human can no longer meaningfully alter the execution under review. Later recovery or response may remain possible.
02Observer Lag+
The gap between executional change and the observer’s ability to receive, stabilize, and act upon a legible account of it.
03Authorization Shadow+
A condition in which formal authorization appears to govern while operative selection or commitment occurs elsewhere.
04Override Theater+
An intervention mechanism presented as control although its timing, access, or practical effect cannot govern the relevant execution.
05Interface Legibility Failure+
The interface does not make the relevant system state sufficiently legible for timely, meaningful intervention.
06Aegis Gap+
The interval between claimed human governance and demonstrated human capacity to affect the execution before material commitment.
07Governance Findings+
Supported, Limited, Unsupported, Shadowed, or Misrouted Governance. A finding remains specific to the claim and boundary reviewed.
PRIMARY PUBLICATION
AEGIS: Post-Execution Trace Governance for Automated Embodied Systems
Observer Lag, Authorization Shadow, and Responsibility Routing Under Compressed Human-Machine Execution
Paper version 3 · KCWorks record version 4. The broader research program includes ten DOI publications and a public implementation repository.AEGIS tests whether a defined human-governance claim is structurally supported. It does not certify safety, decide liability, or replace safety engineering, legal review, incident investigation, compliance, human-factors testing, or conventional risk assessment.