Marcomony in the Physical sciences

In the physical sciences, Marcomony looks a lot more like parsimony. This is because primary utility is what we are tracking back to at all times, and there is little that the physical sciences rely on that is considered "synthesis". Parsimony is strictly the reduction of parameters that give the same quantitative answers to questions.

If one were to do a Marconomic analysis on the Maxwell equations, or Newtonian mechanics, or the General Theory of Relativity, there is very little secondary utility to track back from. One can say that the challenges to Newtonian Mechanics came from new information, such as the inaccuracies in the orbit of Uranus and Mercury. This is not anything to do with Marcomony. However, new information can enable pieces to be placed together with track backs from secondary utility to give strength to alternative disposable razors in cases where a paradigm relies on secondary utility.

Secondary utility can easily creep into models built from the ground up based on reliable physics principles. For example, Climate studies start with the models that reliably predict the weather days in advance. However, the climate models are extrapolated far beyond their capacity to demonstrate primary utility. 

As a general rule, whether it be weather, evolution, economy or any other science that is not strictly physical in nature, simplifications (or parsimony) are required in the transition from a model done from the ground up with physical laws, to a paradigm outside of the realm of physics. Thus, whether it be a climate model, or a comet model, or a model of mutation in evolutionary science, a track back to primary utility is crucial.

In these cases, the confidence these models have gained are from their roots in the physical science. Simplifications are made, and parameters are adjusted to "predict the past" or the term I like to use is "hindcast". Without these models reliably predicting the future, their whole basis is secondary utility. The key to progress is variety of models, different ways of simplifying and therefore different quantity and style of parameters - ie. Alternative paradigms hindcast into historical data rather than adjusting the parameters on a "chosen" paradigm.

Learning robots of the future can only take on the progress of science with this kind of algorithm. Our brains can do this to some extent, and subconsciously, some of us do that, but it is a process that may be able to be automated to be more thorough.