AIM Framework glossary
Architectural Absence Test
Canonical definition
The Architectural Absence Test is the existence-proof-by-absence diagnostic that uses comparator systems lacking specific AIM-source architectures to isolate the source-specific contribution to behavioural output. The general form of the test is: where a system processes the same informational content as another system but lacks the architectural conditions that instantiate a given AIM source, behavioural divergences between the two systems on that content isolate the source's contribution; behavioural convergences falsify the source's contribution to the convergent pattern. The test was originally formulated for M (using AI as the M-absent comparator, per the Mimetic-Contagion-Free Comparator construct) and is generalised by the AI Null-Hypothesis Comparator Principle to all three AIM sources, with AI serving as the simultaneous A-absent, I-absent, and M-absent comparator.
Explanation
The cleanest way to prove a thing is doing work is to remove it and see what changes. The Architectural Absence Test does this for AIM sources by comparing human behaviour to AI behaviour on the same content. AI lacks the bodily homeostasis required for A, the autonomous prior-maintenance required for I, and the biological transmission channel required for M. So whenever humans and AI behave the same way on a task, that behaviour cannot be evidence that A, I, or M is at work; whenever they diverge, the divergence is a candidate signature of A, I, or M contribution. The original M-specific version of this test established that the biological transmission channel, not informational content, is what generates Mimetic Desire. The generalised version extends the same logic to A and I. Extension: The M-specific instantiation remains the architecturally critical case because of the Camouflage Architecture, AI processes M-contaminated content without generating new M-contagion, and AI Confabulation lacks the directional bias toward M-concealment that human Confabulation exhibits. This asymmetric output (sincere wrong output in both, M-directional concealment in humans only) was the original empirical anchor of the test. The generalised form predicts equivalent diagnostic asymmetries for A and I: AI does not exhibit deficit-cyclicity in its output (no A-satiation pattern), and AI does not exhibit autonomous frontier-generating engagement when isolated from prompts (no I-deepening pattern). Each absence is diagnostic of the corresponding human source. The Test is foundational to AIM's claim to architectural realism, the sources are not theoretical posits but architectural conditions whose absence in a comparator system produces measurable behavioural divergence.
Epistemic status
Framework-Specific Named Result.
Derivation
Also in the entry
Source: AI Null-Hypothesis Comparator Principle; Axioms 1, 2, 5, 7.