Explorer, visualiser, décider : un paradigme méthodologique pour la production de connaissances à partir des big data

Abstract : The aim of this Ph.D. thesis in philosophy of science was to answer a practical problem that arose while I was working as an analyst for a software company: how to produce valid knowledge by manipulating large amounts of data that I did not create, and that are not the result of a recognized scientific method? Using my field experience, I propose a methodological paradigm for the construction, exploration and interpretation of Big Data. By methodological paradigm, I mean a theoretical and practical framework that provides keys to developing a method adapted to the data and epistemic project under consideration. By drawing a distinction between the myth of big data and actual practices, I show how digital data are technically and epistemologically constructed from the footprints ("traces") left by individuals. This construction is based on a logic of constitution that requires an interpretative framework and an epistemic continuity between the data and the knowledge we seek to produce. The cultural sciences ("sciences de la culture") thus provide a necessary, but not sufficient, framework to ensure this continuity. Computation, embodied by data sciences and artificial intelligence, materializes and instruments this continuity at the cost of renouncing to be considered as an end in itself. The mediations of interpretation, narrative and software design materialize, show and contextualize the knowledge thus produced. Finally, this knowledge is legitimized by its ability to propose or facilitate decisions and actions.
Complete list of metadatas

Cited literature [193 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01960545
Contributor : Eglantine Schmitt <>
Submitted on : Wednesday, December 19, 2018 - 2:28:07 PM
Last modification on : Tuesday, January 8, 2019 - 1:22:05 AM
Long-term archiving on: Wednesday, March 20, 2019 - 9:46:12 PM

File

181218 - Texte complet - HD.pd...
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01960545, version 1

Citation

Eglantine Schmitt. Explorer, visualiser, décider : un paradigme méthodologique pour la production de connaissances à partir des big data. Histoire, Philosophie et Sociologie des sciences. Université de technologie de Compiègne, 2018. Français. ⟨tel-01960545⟩

Share

Metrics

Record views

1608

Files downloads

1064