Bremen 2017 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

AGPhil: Arbeitsgruppe Philosophie der Physik

AGPhil 6: Symposium: Epistemology of Big Data in Physics IV

AGPhil 6.1: Vortrag

Freitag, 17. März 2017, 10:30–11:00, GW2 B2900

Data science and explanatory power — •Sergey Titov — Institute of Philosophy, Russian Academy of Sciences, Moscow

The analysis-available data has grown immensely in the past decades, it has lead us to the new type of research called "data-intensive". This research design mostly relies on vast amounts of data and use of complicated (commonly non-parametric) statistics. Such data analysis techniques show impressive results in predicting phenomena or it*s characteristics (for example, climat models) but suffer from serious loss in explanatory power (Calude & Longo, 2015). In some cases, models which are generated by nonparametric methods on given data are so complex, that it is nearly impossible to understand its structure. This problem is contemplated from philosophical and mathematical points of view. From philosophy's standpoint authors provide new structures of science which either take data-intensive research into account (Pietsch, 2015) or are fully based on it (Napoletani & Panza, 2011). Second approach attempts to explore mathematical foundations of this loss in explanation power. Calude and Longo in their work (Calude & Longo, 2015) use Ramsey theory and demonstrate that some of the patterns found in data-intesive research may be caused only by size of dataset and nothing else.

This work gives all-round view on this problem and tries to analyse some of the data-intensive researches in the manner described above.