Big Data, Search Algorithms and Philosophical Inquiry

On January 18, 2019, Clark Glymour, Joseph D. Ramsey and Kun Zhang have published an article entitled “The Evaluation of Discovery: Models, Simulation and Search through “Big Data”” in the Open Access journal Open Philosophy.
A Blog Post by Pablo Markin.
In their paper, in reaction to the constant growth in the big data, such as in the domains of climate science, neuropsychology, and astronomy, Glymour, Ramsey and Zhang have inquired into the accuracy of the corresponding data analytics and simulation algorithms, in order to “show that the model/simulation strategy is not confined to causal inference” (39), on the basis of “classification problems from astrophysics” (39). Consequently, this paper revolves around the following questions: “Would a “precise method of discovery” […] be only an algorithmic procedure guaranteed to produce plausible or probable explanations? Was a single method envisioned that would be applicable to every domain, or as Rawls’ plural “methods” suggests, possibly distinct methods for different branches of inquiry?” (40).
As these researchers highlight, “[a]lgorithmic methods for searching for relationships that indicate causes or allow the identification and classification of phenomena were proposed early in the 20th century, notably in factor analysis. They lacked any demonstrations that they are truth guiding even in the infinite limit” (41). Furthermore, Glymour, Ramsey and Zhang note that “[w]hether causal information can be extracted for an entire body of data depends on the dimensionality of the data (measured in number of variables) and the complexity of the data generating system (measured in terms of the density of connections in the unknown, true causal graph), on the sample size, and on the speed of the algorithms” (43).
https://www.instagram.com/p/BwrzeUknipT/?utm_source=ig_web_button_share_sheet
Thus, these authors conclude that “the ambition to find a single method for all problems of empirical inquiry has not been realized” (47). According to Glymour, Ramsey and Zhang, “[m]odel simulation done with sufficient care and attention to variety, supplemented with other evidence, can provide a basis for confidence in “precisely describable” methods of discovery. But procedures for evaluating search methods by simulations are themselves often not automated in selection of examples, and are without guarantees that they are representative. Developing formal, sensible measures of confidence in such circumstances is an interesting challenge, and that may be where “precision” really is impossible” (47).
By Pablo Markin
Featured Image Credits: New Simulation Sheds Light on Spiraling Supermassive Black Holes, October 3, 2018 | © Courtesy of NASA’s Goddard Space Flight Center/Flickr.
OpenEdition suggests that you cite this post as follows:
Pablo Markin (April 16, 2019). Big Data, Search Algorithms and Philosophical Inquiry. The DGO Blog. Retrieved April 28, 2025 from https://doi.org/10.58079/nkcf