Because data of interest is usually, if not always, produced by mechanistic and algorithmic causes, I have proposed an empirical Bayesian approach to science where uninformative priors are not uniform flat (said uninformative) distributions but Levin’s so-called Universal distribution based on algorithmic probability (Solomonoff). I am thus introducing different tools and methods than those based on (statistical) machine learning and Shannon entropy which are more likely to produce false models from spurious correlations specially in the advent of Big Data. To this end I am introducing algorithmic-information tools better equipped to model and deal with the discovery of first and design principles and generating mechanisms (causes). See papers: J18, J20, J26, J27, J31, J31, J32 and P29, P30, P26.
