Assessing biases, relaxing moralism: On ground-truthing practices in machine learning design and application

Abstract

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: (I) the definition of the problem to be solved (problematization), (II) the identification of the data to be collected and set up (databasing), and (III) the qualification of the targets to be learned (labeling). 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.

Hybrid Scientists: Agency, AI, and the Science of Data

“…A good machine learner is assessed based on performance metrics based on evaluations against a particular dataset-what is often called a ground truth dataset (a term borrowed from meteorology) (Krig, 2014a;Jaton, 2017Jaton, , 2021. Ground truth datasets are sometimes described as the true measurements of what you want to predict.…”

Tradeoffs all the way down: Ethical abduction as a decision-making process for data-intensive technology development

“…In observing how the AMOS team viewed disability as a social construct arising from the interplay between cultural assumptions and technological design, how this understanding informed their recognition of the ethical implications of the data infrastructure they were designing, and how this worldview manifested in the pragmatic choices they had to make, we can see that ethical abduction entails tacking back and forth between divergent but complementary ways of thinking: between establishing theoretical ideals and making decisions given practical constraints; between understanding historical context and anticipating future consequences; between acknowledging structural dependencies and accepting responsibility for moral agency. Other scholars have attended to the mundane, ongoing, situated nature of ethical decision-making in data-intensive practices (Jaton, 2021; Leonelli, 2016; Metcalf et al, 2019; Rességuier and Rodrigues, 2020; Ziewitz, 2019). The purpose in naming, defining, and describing “ethical abduction” as a particular mode of situated ethical practice is to provide a precedent that is concrete enough to be emulated by practitioners of data-intensive technologies and the architects of the environments in which they work—and yet flexible enough to be adapted to unique contexts of production.…”

“…It marks a shift away from treating ethics as “a form of argument or an abstraction” and toward the development of an “ordinary ethics” (Metcalf et al, 2019: 455), which, in keeping with the analytical philosophy of Wittgenstein (cited in Lambek, 2010), treats ethics instead as “moral commitments embedded in actions” (Metcalf et al, 2019: 455). Like Jaton (2021), I attend to moments of “hesitation” (p. 2) when actors face what pragmatist philosopher William James called “genuine options” (cited in Jaton, 2021) offering up distinctly divergent futures. Noticing such consequential junctures requires the “attentiveness” called for in feminist ethics of care (Tronto, cited in Rességuier and Rodrigues, 2020: 3) that can result in a “constantly refreshed capacity to perceive the world” (Rességuier and Rodrigues, 2020: 3).…”

Strengthening Equity and Inclusion in Urban Greenspace: Interrogating the Moral Management & Policing of 2SLGBTQ+ Communities in Toronto Parks

“…Ground-truthing, through on-the-ground observations of space, has become commonly applied in broader geographic and social research to verify data in everyday contexts, and understand biases and perspectives presented as truth [ 41 ]. Instances of in/exclusion or othering in greenspace were examined through field observations of symbolic and concrete codification of prescribed desirable/undesirable identities and behaviours of greenspace users.…”