Data-intensive methods and technologies hold great promise, but also have the potential to produce misleading results [25], violate rights and norms [35], perpetuate structural disparities [5], and inhibit accountability [22]. Accordingly, in recent years we have seen a burgeoning response in the form of methodology [40], methods critique [28, 29], pedagogy [43], principles [11], and projects [21] aimed at rehabilitating data-intensive technologies and methods to make them more ethical, responsible, equitable, and beneficial. Much of this discourse either explicitly or implicitly advocates for making data-intensive methods and related technologies more “human-centered” [36] by designing for “human values” [8], considering “human contexts” [42], and prioritizing “human flourishing”[39].
Human-centered perspectives offer powerful antidotes to beliefs and practices that naively assume the beneficence and superiority of technology or pursue technological advancement regardless of harms inflicted upon people in the process—a phenomenon that Broussard [9] has called “technochauvinism.” However, the HCI analyses (e.g., [3, 20, 26, 27]) have not thoroughly examined their own grounds of criticism. In this workshop, we deepen that critical view by turning a reflective lens on the HCI work itself that addresses data science.
Scholars have questioned the ability of anthropocentrism to adequately address the moral, political, environmental, and social challenges posed by sociotechnical systems. This includes perspectives of feminist and Indigenous epistemologists who focus on the web of relationships between humans and nonhuman entities rather than on the human as the preeminent location of attention. Such works break down simplistic ontologies of difference and explore how humans are constituted in relation to nonhumans (Haraway, 2016/1985), reminding us that “human life is embedded in a material world of great complexity, one on which we depend for our continued survival” [18, p. 5].
These observations seem particularly salient given recent revelations about the ecological impact of data-intensive methods and technologies [4, 10], as well as moral questions raised by the possibility of sentient artificial intelligence [41]. Lewis et al. [23] argue for a position on AI drawn from Indigenous ontologies that “take the world as the interconnected whole that it is” in order to “illuminate the full scale of relationships that sustain us, provide guidance on recognizing non-human beings and building relationships with them founded on respect and reciprocity.” And HCI researchers have begun exploring ways that decentering the human may open new avenues for equitable and sustainable technology design [6, 15, 38].
Such ideas point to multiple ways in which relational thinking may inform our approaches to interrogating data science. Sabina Leonelli [22] urges us to recognize data as “relational” and “fungible” objects that are “first and foremost, material artifacts” whose “physical characteristics, including their format and the medium through which they are conveyed, are as relevant to understanding their epistemic role as their social and conceptual functions” (for treatments of data as a medium of design, see [12, 13, 32–34]). Birhane and Cummins [7] call attention to the ways in which it is problematic to assume that technologies may ever be rendered beneficial to an undifferentiated set of all humans or achieve a common good, advocating instead for a “relational ethics” that prioritizes an algorithmic technology’s impact on the most vulnerable, marginalized, or disproportionately affected.
In this workshop, we invite explorations of these and other provocations salient to “human-centeredness” as it relates to data science. Previous events convened by the organizers of this workshop and other colleagues [3, 20, 26, 27] under the guise of human-centered data science have critically interrogated data science and generated practical approaches to mitigating the social and ethical challenges posed by the field. These discussions led to the writing and publication of the book, Human-Centered Data Science: An Introduction [2], forthcoming from MIT Press. We leverage the publication of that textbook as an occasion to probe and perturb conversations and practices around human-centered data science—including the conversations that led to the book. Whereas the aforementioned workshops were tasked with “interrogating data science,” here we turn our lens back on that interrogation itself, creating an opportunity for reflection that advances the field of human-centered data science in new directions.