Human-in-the-Loop architectures require active human authorization before a consequential decision executes; Dead-Hand architectures pre-authorize decisions to execute automatically. The distinction looks clean in the abstract, but the project's argument is that it collapses in practice: interface design, time pressure, cognitive load, and organizational incentive routinely make meaningful human review impossible, even where it remains nominally required.

The Delegation Architecture framework is the project's central contribution — a methodology for testing whether human authority in an autonomous decision system is real or procedural, across four dimensions: cognitive feasibility of review, organizational accountability for outcomes, the interface's actual (versus nominal) affordances for intervention, and the incentive structure governing whether operators exercise or waive their authorization role.

The framework's primary application is military autonomous systems, where the Human-in-the-Loop requirement is often mandated by policy or treaty while being eroded by operational tempo. It applies equally to civilian high-stakes domains — medical diagnosis, judicial risk assessment, financial intervention — wherever the gap between nominal and actual human authority creates accountability voids.

The research is mature: a Zenodo publication establishes the framework, an essay develops the argument, and a lecture is published in English.