4-Slide Mini Deck · Beyond Performance
Beyond Performance, in four slides
Micheal Charles Preble · SSRN 7416901 · Pre-validation diagnostic method
Slide 1 of 4
The Problem
Claims that AI improved human performance increasingly appear in education, professional work, and public administration — but improvement in an AI-mediated system does not by itself establish that the human participant advanced.
Slide 2 of 4
The Beyond Performance Contribution
A diagnostic method built on three configurations (autonomous capability, governed extended capability, ungoverned dependency) and three patterns (coupled improvement, performance-only acceleration, regressive decoupling), returning a succeed / unproven / fail disposition for a declared claim.
Slide 3 of 4
How the Framework Works · Evidence Status
Applied to two published third-party studies: Bastani et al. (2025) — standard GPT-4 assistance: +48% practice, −17% unassisted exam; guarded tutor removed the penalty without a learning gain. Strömberg et al. (2026) — +18% homework score, −20% exam decline, concentrated among an outsourcing-majority group.
Evidence status: Pre-validation decision procedure, not a validated instrument.
Slide 4 of 4
Why It Matters · Limits · Open Questions
Stronger outputs can coexist with losses in capacities necessary for meaningful agency. The method forces that distinction to be tested rather than assumed.
Limits: not a validated instrument; depends on predeclaration; inherits the limits of the evidence evaluated. Open: cross-domain validation; independent-evaluator consistency.
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.