A must-attend event for young researchers in the philosophy of science, the Society of Philosophy of Science organizes annual International Doctoral Degrees in Philosophy of Science (Rencontres doctorales internationales en philosophie des sciences, RDIPS in French).
For this 7th edition, these meetings will take place from Wednesday, September 26 to Friday, September 28, at the University Paris 7 Paris Diderot (Thesis Room 580F, La Halle aux Farines).
Eglantine Schmitt, our Product Manager, is here for the third time. This year she will have the honor to close the first day, Wednesday, September 26, 2018 from 16:20 to 17:00.
What kind of knowledge do algorithmic recommendation systems produce?
Data sciences are considered as a set of epistemic practices that aim to manipulate digital data, usually present in large volumes, to produce new knowledge. Unlike physics, biology, or sociology, which deal with an object of the world (nature, living, society …), these data sciences are rather a set of techniques that can be applied to various data, which can designate objects of the world also very diverse. In particular, they can be mobilized in a scientific and university context, but also in the service of professional and commercial practices for which the production of knowledge is not systematically an end in itself.
In this perspective, we wish to show:
- That the data sciences are not strictly speaking a science in the classical sense, but a technical and methodological toolbox that can be mobilized in different scientific or epistemic contexts;
- That the digital substratum of the production of knowledge in the data sciences makes possible two forms of restitution of this knowledge: a narrative or a computational system;
- That in the second case, this system is almost always an algorithmic recommendation system whose function is not to clarify knowledge but to provoke an action;
- That the immense majority of the techniques used in the data sciences (hierarchization, classification, grouping) lend themselves to the construction of a recommendation algorithm;
- That the recommendation system is a form of reproduction that adequately describes the services offered by most major digital players (Google, Amazon, Facebook, Twitter, Netflix …);
- That the action evoked by such a system is made possible, but also delimited, by said system, consisting of the recommendation algorithm itself, but also of its digital environment;
- That by arousing an action, a system of recommendation produces knowledge of a practical nature, even pragmatist, that is to say, knowledge which is knowledge only because it makes possible and aims at making possible an action.
While it is true that Peirce’s pragmatism can also be applied to scientific and academic knowledge, the relationship between knowledge and action is weaker than in recommendation systems whose aim is specifically to generate a decision for action. We therefore propose to consider this relationship between knowledge and action as the defining demarcation criterion for characterizing, on the one hand, the knowledge produced in a scientific framework, and on the other hand, the knowledge produced by the recommendation algorithms.
More information: Society of Philosophy of Science