Ashby had powerful insights. The Good Regulator theorem of Conant and Ashby that every good regulator of a system must be a model of that system is instructive. Further down the Ashby path – the variety (a.k.a. complexity) of a model needs to be at least as sophisticated as that of the system it models.
Of course, there are (very) many paths to usefully compress system complexity for both models and explanations of those models. The center of gravity here sits somewhat outside the systems models themselves and in the space of exegesis and education that envelops and infuses them.
The beauty (and the burden) of scientific models is that they constructively diminish, limit or damp possibilities. This compression is not coincidentally the core of intelligence and – indeed – of living systems.
A generalised error we (all) make here is to assume or assert the preeminence and ascendancy of the models (as component microcosms) over the systems they describe when the underlying, unifying ontology is significantly more subtle and sophisticated.
Reductive models amplify effect but displace entropy into themselves and their environments as ambiguity and uncertainty. Offsetting complexity as shaped entropy is a way out of this tesseract.