Concept Explainer · Beyond Performance

The diagnostic method, explained

A concise reference for readers who want the core distinctions without reading the full paper.

What are the three configurations?

Autonomous capability — a function the person can perform without live external assistance. Governed extended capability — a function materially expanded by AI while the person retains the capacities and practical powers to direct, interpret, contest, and sustain the arrangement. Ungoverned dependency — high or improved functioning that depends on a system the person cannot meaningfully evaluate, challenge, substitute, or recover from where those powers remain necessary.

What are the three diagnostic patterns?

Coupled improvement — AI-mediated performance rises and the relevant human-side conditions improve or stay robust. Performance-only acceleration — output improves but human-side evidence is flat, absent, or insufficient. Regressive decoupling — AI-mediated performance rises while a declared human-side condition materially deteriorates.

Why does the distinction matter?

Most AI evaluation stops at output: accuracy, speed, completion. Those measures are necessary but insufficient once the conclusion shifts from a claim about the system to a claim about the person. An AI-mediated configuration can improve while the human participant becomes more capable, less capable, or differently capable in ways the artifact alone cannot reveal.

What is actually claimed?

  • That a progress claim must declare, before evaluation, the human functioning at issue, the population and role, the scope, the required contrast, the durability window, and the failure signs.
  • That observed change can be classified into the three patterns above.
  • That, applied to two published third-party studies, an assisted-performance claim can succeed while a stronger human-learning claim fails or remains unproven.

What remains unproven?

  • Whether independent evaluators applying the same declared claim and failure criteria reach the same dispositions.
  • Whether the method generalizes to domains beyond education, such as public administration or professional work.
  • The domain-specific thresholds that separate material from immaterial deterioration.

Where can readers go deeper?

The canonical paper record · Research brief · Decision-table research map · Human Capacity

Suggested citation
Preble, Micheal Charles. “Beyond Performance: A Diagnostic Method for Evaluating Human-Advancement Claims in AI-Mediated Systems” Manuscript v2.3. Available at SSRN 7416901, 2026.