AI alignment relies on discredited notion of universal humanity that is the root cause of mass atrocity

(Daniel Estrada, Ethics@Newark, Human supremacy as posthuman risk. Journal of Sociotechnical Critique, 1(1), 1–40.

https://digitalcommons.odu.edu/cgi/viewcontent.cgi?article=1013&context=sociotechnicalcritique)

Against the backdrop of numerous political scandals, ethical violations, and calls for regulatory oversight in the field of artificial intelligence (Whittaker et al., 2018), the rhetorical framework of the “human” has become an increasingly visible shorthand for industry and public policy projects to signal a concern for safety, ethical integrity, and the responsible use of the AI. Several recent public policy proposals on AI bear titles such as “AI for Humanity” in France; 1 “HumaneAI” in the EU; 2 “AI4People”, Luciano Floridi’s proposal describing “An ethical framework for a good AI society” (Floridi et al., 2018), and Stanford’s “Institute for Human-Centered AI,” whose welcome page proudly proposes that “AI is to serve the collective needs of humanity.”3 In these contexts, centering the “human” as the explicit focus of normative concern projects the appearance of an inclusive framework of shared values and common interests that guide the collective use of AI. More subtly, these proposals consistently frame AI as categorically subservient to human interests. Unsurprisingly, these proposals don’t define the scope or content of the “human,” how membership in this category is to be determined, or what underpins its role as a focal norm in AI policy. Still, AI is regularly used to target, manipulate, incarcerate, and exploit vulnerable populations, as seen in the use of facial recognition technologies in policing and military operations, election tampering and voter manipulation on social media, and software that automates criminal sentencing, loan approval, hiring decisions, and so on (Williams, 2019b; Spiel et al., 2019; Buolamwini & Gebru, 2018; Eubanks, 2018; Noble, 2018; Asaro, 2016; Angwin et al., 2016; Benjamin, 2016). Given these realities, how can we trust that policies which center the “human” will also center us? It is difficult to see how technologies used to expand, reinforce, and make more profitable these pervasive institutions of violence and oppression could operate against a shared background of values common to all humanity. Rosi Braidotti quotes Tony Davies: “All humanisms, until now, have been imperial” (Braidotti, 2013, p. 15). From this perspective, the attempt to signal a commitment to integrity by appeal to the category “human” reads as more of the empty ethics-washing that has come to characterize the field (Metzinger, 2019). Absent from these human-centered proposals is any engagement with the decades of sustained scholarship in feminist, postcolonial, and critical race theory (Jackson, 2020; Deckha, 2012; Mills, 2011; Hayles, 2008; Said, 2004; D. Haraway, 1989, 1991), animal rights and environmental ethics (Belcourt, 2015; Gaard, 2011; Steiner, 2010; Katz, 2000); STS, HCI, and design theory (Thomas et al., 2017; Kera et al., 2009; Latour, 2003), and related fields that have developed systematic critiques of anthropocentrism and the politics of the “human.” Among the important insights of this diverse literature is the recognition that a superficial appeal to inclusive universalism can be used to justify and provide cover for narrowly self-serving, exclusive, or imperialist practices (Khader, 2018; Giraldo, 2016). As Charles Mills puts the point, “historically and still currently, most humans were not and are not socially recognized persons, or, more neatly and epigrammatically put: most persons are non-persons” (Mills, 2011). Confronting such duplicitous ideologies presents difficult conceptual, rhetorical, and practical challenges, suggesting that care should be taken in the use of universalizing language if it is used at all. The uncritical deployment in AI policy of human-centered rhetoric as a pretense to ethical integrity speaks not only as a tacit endorsement of its imperialist undertones, but more loudly as utter disregard for those scholars and activists who have been consistently engaged with the “human” as a normative ground.