This theoretical paper considers the morality of machine learning algorithms and systems in the light of the biases that ground their correctness. It begins by presenting biases not as a priori negative entities but as contingent external referents—often gathered in benchmarked repositories called ground-truth datasets—that define what needs to be learned and allow for performance measures. I then argue that ground-truth datasets and their concomitant practices—that fundamentally involve establishing biases to enable learning procedures—can be described by their respective morality, here (...) defined as the more or less accounted experience of hesitation when faced with what pragmatist philosopher William James called “genuine options”—that is, choices to be made in the heat of the moment that engage different possible futures. I then stress three constitutive dimensions of this pragmatist morality, as far as ground-truthing practices are concerned: the definition of the problem to be solved, the identification of the data to be collected and set up, and the qualification of the targets to be learned. I finally suggest that this three-dimensional conceptual space can be used to map machine learning algorithmic projects in terms of the morality of their respective and constitutive ground-truthing practices. Such techno-moral graphs may, in turn, serve as equipment for greater governance of machine learning algorithms and systems. (shrink)

Ethnographic study of the constitution of algorithms.

In this paper, we build upon Bruno Latour’s political writings to address the current impasse regarding algorithms in public life. We assert that the increasing difficulties at governing algorithms—be they qualified as “machine learning,” “big data,” or “artificial intelligence”—can be related to their current ontological thinness: deriving from constricted views on theoretical practices, algorithms’ standard definition as problem-solving computerized methods provides poor grips for affective dissensions. We then emphasize on the role historical and ethnographic studies of algorithms can potentially play (...) in the politicization of algorithms. By both digging into the genealogy of algorithms’ constricted definition and by making their contemporary constitutive relationships more visible, both historical and ethnographic studies can contribute to vascularizing algorithms and making them objects of enlarged disputes. We conclude by giving a flavor of the political potential of the vascularization efforts we call for, using materials from an ethnographic study conducted in a computer science laboratory. (shrink)

S’appuyant sur trois enquêtes de terrain qui portent sur le façonnage d’êtres numériques, le présent article traite des relectures méthodologiques que l’on peut faire de trois contributions de Latour à l’enquête ethnographique. 1. Il montre que l’accent mis sur les séquences d’action permet d’engager une enquête systématique qui rend compte de trajectoires. 2. Il avance que le concept d’actant sensibilise à l’éventail des entités agissantes qui infléchissent ces trajectoires, de même qu’au peuplement de situations apparemment inoccupées. 3. Il soutient que (...) les concepts d’association, de traduction et de médiation permettent de connecter différentes situations et actions localisées. Ultimement, ces trois apports éclairent les agencements et les réagencements.Rooted in three fieldwork studies that examine the shaping of digital information, this article discusses new methodological readings of three of Latour’s contributions to ethnographic inquiry. 1. It shows that focusing on sequences of action allows one to engage in a systematic investigation that accounts for trajectories. 2. It proposes that the concept of actant makes one aware of the whole set of active entities that influence these trajectories on the one hand, and the populating of apparently empty situations on the other. 3. It maintains that the concepts of association, translation, and mediation make it possible to connect various localized situations and actions. Ultimately, these three contributions shed light on arrangements and rearrangements. (shrink)