The Validity of Occam's Razor

The philosophical spirit of ancient Chinese medicine is compared here with the famous principle of Occam’s Razor in machine learning and philosophy of science. Its central idea is: entities should not be multiplied without necessity. When constructing models to explain the world, if two models achieve the same predictive or therapeutic effect, the model with the simpler structure and fewer rules will usually be closer to objective truth. In this framing, the generalization ability of a model is proportional to the parsimony of its rules. Models that are excessively complicated often fall into what machine learning calls overfitting: they can explain special cases that have already occurred, but cannot predict future change. When ancient Chinese medicine confronted complex phenomena, it recognized that regulating a complex system first requires capturing the laws of the energetic drivers behind it. If the information engineering of knowledge compression were pushed to its limit across the tens of thousands of Chinese characters in the Huangdi Neijing, what would remain would be only several dozen core symbols representing different energy fields—Yin and Yang, the Five Phases, the Heavenly Stems, and the Earthly Branches—together forming a highly compressed model of interaction between Heaven and humanity. In this view, that is an ultimate expression of Occam’s Razor. A model of Chinese medicine assembled from first principles should neither have rules removed beyond what is necessary nor have arbitrary rules added. Just as Copernicus’s heliocentric model, with its simpler orbital conception, ultimately displaced the far more complicated Ptolemaic geocentric system, parsimonious models tend to prevail when confronting complex systems.