[2J4-GS-1] Fundamental AI, theory: algorithm

[2J4-GS-1-05] On cognitive model selection criteria that prefers the Heliocentrism over the Geocentrism.

〇Natsuko Katase¹, Takuma Torii², Shohei Hidaka¹ (1. Japan Advanced Institute of Science and Technology, 2. Tokyo Denki University )

Keywords:Planetary motion , Geocentric model , Heliocentric model , explainability , predictive error

If we can learn the method of discovering physical laws, the speed of finding the physical law would be boosted. In cognitive science, past research has attempted to understand the framework of sciences using analogies. In computational science, past research tried to find physical laws using machine learning. However, the former is not based on data but on concepts. The latter is modeling based on prediction errors and has no criteria other than prediction errors. In this study, we consider a problem with the prediction of planetary motions using geocentrism or heliocentrism models. Copernicus' model based on the heliocentrism model has a bigger error than Ptolemy's model based on the geocentrism model, but humans selected Copernicus' model based on history. There is another criterion aside from error. The average bias in Ptolemy's model increased in the retrograde motion parts more than in other parts.