Abstraction Jumps in Medical Knowledge: Traditional Chinese Medicine, Cognitive Constraints, and AI-Enabled Intermediate Abstraction

Traditional Chinese medicine (TCM) and modern biomedicine describe the
same human organism through different architectures of abstraction. This
paper argues that an important part of the divergence lies not in
whether the two traditions abstract, nor simply in how abstract their
concepts are, but in the density and structure of the representational
paths connecting clinical observations to higher-order medical objects.
TCM contains highly abstract concepts and diagnostic categories, yet in
important domains these have historically been connected to
phenomenological observations through comparatively large
representational distances. We call such cases abstraction jumps:
transitions in which a cognitive system compresses a complex lower-order
field into a substantially higher-order object without stabilizing a
sufficiently articulated sequence of intermediate representations.

The paper develops three further ideas. First, abstraction
infrastructure---measurement, intervention, representation, and
validation---lowers the cost of constructing and stabilizing
intermediate abstractions. Second, abstraction-ladder density
characterizes the density of stable, manipulable, transferable, and
independently testable representations between observation and
higher-order theory. Third, artificial intelligence can alter the point
at which compression becomes necessary. When machine systems can sustain
relational complexity beyond the effective human threshold for a task,
they may permit delayed compression and the discovery of previously
unstabilized intermediate structures.

Syndrome-element theory is interpreted as an existing form of endogenous
densification within TCM rather than as evidence that TCM lacks
intermediate structure. AI extends this developmental possibility only
if machine-discovered structures survive independent qualification
through stability, replication, transfer, cross-readout comparison,
incremental epistemic value, and, where appropriate, prospective or
intervention-sensitive testing. The resulting proposal is neither
biomedical assimilation nor static preservation. It is a third path in
which a medical knowledge system becomes more internally articulated and
empirically revisable by generating new intermediate objects from within
its evolving empirical domain.