What It Studies
The limits of detection itself. AMPL proved that cognitive pressure leaves a measurable trace in voice. TennisMiner showed that motor breakdown under pressure has observable precursors. The next step seemed obvious: detect the signal, alert the person, prevent the breakdown.
That is not what the research found.
Why It Matters
Every monitoring system — whether for pilots, CEOs, athletes, or AI oversight operators — assumes that if you can detect the problem, you can intervene. But detection without understanding its structural limits creates false confidence. An alarm that cannot tell you when it stops being reliable is more dangerous than no alarm.
What We Found
Two formal results constrain the problem:
External verification is recursively unstable. Any finite verification chain depends on at least one unverifiable assumption. An instrument that says “this person is under cognitive pressure” requires someone to interpret the reading — and that interpreter’s own cognitive state cannot be verified by the same instrument without regress. This is not an engineering limitation. It is a structural property of the verification architecture.
Internal detection is subject to imperceptible drift. An observer’s baseline can be shifted arbitrarily far through sub-threshold incremental steps. Each step is below the perceptibility threshold; the total drift can be unbounded. A person under gradually increasing pressure cannot reliably detect that their own judgment has been compromised — because the instrument they would use to check is the thing being distorted.
Together, these results close two doors: external monitoring has a ceiling, and internal self-assessment is unreliable.
A third mechanism exists. We present formal and empirical arguments that when a cognitive system models itself with sufficient fidelity, compulsive attachment to specific goals loses its structural support. The system does not need to be told it is breaking. The conditions that sustain the breaking pattern dissolve from within.
The empirical evidence comes from a 4th-century BCE contemplative corpus: 30 mechanism-level claims about cognitive architecture, 26 independently verified by peer-reviewed findings across neuroscience, thermodynamics, information theory, control theory, and evolutionary biology. A negative control comparison against a contemporaneous Western framework (Aristotle’s De Anima) yields 0/10 structurally precise correspondences.
Implication for AMPL
AMPL is a necessary first step — proving the signal exists and is detectable. But the theory constrains what detection alone can achieve. The deeper goal is not a better alarm. It is understanding the conditions under which the alarm becomes unnecessary.
Publication
Phase Transitions in Self-Modeling Cognitive Systems: When Goal Persistence Becomes Structurally Unsustainable Ziyan (Ben) Zhou — Expeta Technologies Inc. PhilArchive, 2026