The public release of OpenAI’s generative artificial intelligence (GenAI) chatbot ChatGPT in November 2022 was, in the jargon of political science, a “critical juncture” in which AI’s potential was made manifest to ordinary people. Following the initial enthusiasm, however, came a strong undercurrent of skepticism about the human costs of AI advances. In 1958, Hannah Arendt presciently observed:

The wish Arendt is referring to here is the wish to be free from the burden of labor and necessity. It is “self-defeating” because, in a society composed of individuals conditioned primarily as laborers, mass unemployment due to automation fosters a kind of dangerous nihilism. AI skeptics often share Arendt’s fear, while also highlighting other, more immediate, concerns about agency, structural injustice, and inequality. It is to these issues that Leslie Paul Thiele’s Human Agency, Artificial Intelligence, and the Attention Economy, Jude Browne’s Political Responsibility and Tech Governance, and Colin Koopman’s Data Equals are devoted.

For Thiele, the main threat to human agency—understood as “reflective self-direction” (243)—posed by emerging technologies, including AI, is “human downgrading” (1–3). Thiele divides the downgrading into three main categories of deskilling: cognitive, social, and civic. According to Thiele, “[c]ognitive skills include memory, critical thinking, judgment formation, and decision-making”, while “[s]ocial skills generally include communication, empathy, and the perception of body language and other social cues” (129). Finally, although they are not as clearly delineated as the previous two, civic skills seem to include by implication something like openness to political dialogue, moderation, discernment, and “real-world” political participation, as Thiele charts the decline of civic skills in terms of increasing “ideological rigidity and political polarization,” which, combined with misinformation and “clicktivism,” feeds “the eclipse of common truths, the decay of civic trust, and growing dogmatism” (153). The result is a breeding ground for autocracy (154).

All digital media are implicated in human downgrading, although Thiele is particularly concerned with the role of AI-empowered social media in this process. The bounty of empirical evidence he marshals to support his argument is impressive, if depressing. For example, he reports that “available data implicate digital media in a wide array of psychological ailments, including addiction, antisocial personality disorder, attention-deficit disorder, depression, narcissism, obsessive-compulsive disorder, social phobia, and voyeurism, among others” (189–90).Footnote ² Equally impressive is the wide array of thinkers Thiele draws on, including Arendt, Friedrich Nietzsche, Karl Marx, Martin Heidegger, and Mahatma Gandhi, among many others. The picture he paints is an admittedly bleak one, but he quotes Romain Rolland (the sentiment was made famous by Antonio Gramsci) in counseling ‘pessimism of the intelligence … and optimism of the will’ (24).Footnote ³

Unlike Thiele, Browne is primarily concerned not with individual human “downgrading,” but with structural injustice, which she defines as “negative outcomes experienced by individuals, groups or institutions that are generated by a vast array of complex, multitudinous ‘everyday’ actions of those and other individuals, groups and institutions operating within accepted norms and institutional practices in a given time and space and accumulated at the macro level” (13). This focus does not mean Browne ignores more traceable, liability-based harms, as she acknowledges that AI’s entrenching of stereotypes and biases and dissemination of disinformation include both forms of harm (59). While such concerns are relatively well-known, Browne also highlights the structural harm caused by AI’s promotion of “thoughtlessness” (54, 60). Such thoughtlessness is related to the offloading of cognitive labor and the “[o]pacity” of AI’s workings (61–3). Other possible forms of structural harm that Browne highlights in relation to AI include mass unemployment, increasing precarity for workers, and the advent of uncontrollable AI (63–8). In the case of “Repro-tech” (reproductive and genetic technologies), which increasingly utilizes AI, Browne highlights the way in which the coverage of egg freezing by many large corporations may not only exacerbate existing inequalities but also increase pressure on women who have access to the service to do so (96–7). Such benefits may also diminish support for “comprehensive state subsidized childcare and paid parental leave” (95). Finally, pre-implantation genetic technologies (PGTs) could enable a social shift toward “enhanced children,” which would exacerbate inequality and, again, likely weaken support for social welfare policy efforts (101–3).

As in Thiele’s book, Browne’s work is filled with striking findings. For example, she relays: “A recent survey of 1,031 UK workers under taken by a AI-recruitment company [that] reported that ‘a third of women (33%) and over two-fifths (43%) of men said that it was likely or very likely that their current job role could be replaced by technology or machines (such as smart software, automation or robotics) in the future’” (65–6).Footnote ⁴ Browne is also like Thiele in drawing inspiration from Arendt’s work. In addition to her concept of “thoughtlessness,” Browne contemplates Arendt’s famous assertion that “Where all are guilty, nobody in the last analysis can be judged” (17).Footnote ⁵ Departing from Arendt, Browne turns to Iris Marion Young (her primary interlocutor, who is herself engaging with Arendt) in asking the book’s central question: ‘where nobody is liable, who is responsible?’ (5, 19; emphasis in original).

Like Browne, Koopman grapples with structural dynamics, and, like Thiele, he is concerned with the undemocratic tendencies of emerging technologies, especially including AI. His book is distinct in that its focus is on the ways in which technological hierarchy—specifically in regard to data formats—undermines what he terms “democratic” or “relational” equality, which refers to the ability of people to relate to one another as equals (5, 8–10, 33). Data formats can undermine equality in a number of ways, including “polarization[,] … segmentation on the basis of prediction, competition as incentivized by gamification, declining trust in online interactions[,] online disinformation, and … data-driven personalization” (33). More generally, in their emphasis on separation and exclusion, data hierarchies are shown to have an affinity with authoritarian rather than democratic logic (28–9).

In his innovative examination of data formats, Koopman considers the example of “EdTech” or “educational technology.” Central to EdTech is personalization—amplified by AI—in the form of “adaptive learning, intelligent tutoring, personalized assessment, and learning prediction” (176). While personalized education may not sound like a bad thing, Koopman criticizes it for the way in which separating or “discretiz[ing]” learners promotes inequality (183). According to Koopman, personalized learning undermines relational equality, both directly through individuation and indirectly by exacerbating existing distributive inequalities (193). In the former, learners are discretized by data formats that conceptualize learning as an isolated process. As a result, collaborative skills are deemphasized, while competition is often encouraged (194–200). In terms of the latter mechanism, personalization operates like traditional student tracking as its “heavy reliance on historical data … tends to reproduce those extant distributive inequalities reflected in the historical data it incorporates into its analytics” (202). In making the case for the unique significance of education to democracy, Koopman draws primarily on the philosophical pragmatism of John Dewey (Elizabeth Anderson and W. E. B. Du Bois are also significant influences although many other thinkers are discussed) in arguing that education is a privileged site of acculturation, where equality is widely desired and has been actively sought (171–4). For these reasons, we should be particularly concerned with the undermining of democratic equality in this domain (212–5).

Each of the authors offers distinct, creative solutions to the AI-related problems they identify. Although all admit the importance of political regulation, they do so to differing degrees. Thiele foregrounds a more individualistic approach in the form of “digital distancing,” which “entails developing the dispositions, habits, and skills that enable reflective self-direction and practices of caring” (18). “Like a well-composed paragraph” (235), reflective self-direction requires restraint, but by practicing it, we can become worthy “protagonist[s]” in the narratives of our lives (241). For him, an essential part of the process is community service because “for human beings to flourish, their agency must be deployed in ways that integrate them within a greater whole” (243). In sum, Thiele calls for people to reclaim control of their thoughts and actions—redirecting their energies from the self to the community and from the online to the offline world. He admits this is no small task as it requires us to “actively rewire our brains” (284). Central to the effort is the crafting of a narrative rooted in “self-leadership and service,” which produces new pleasures that can foster a “virtuous cycle within the riptide of an AI-enabled attention economy” (285).

Browne reverses Thiele’s ordering in prioritizing government regulation without neglecting the role of the individual. Her proposal is for a regulatory body for AI based on the example of the UK’s Human Fertilisation and Embryology Authority (HFEA). As a public body, the HFEA is “charged with protecting the public interest” (105). It does so through regulation, licensing, and providing the public and government with information and guidance on the topic (105–7). Its workings are characterized by Browne as “evidence-based deliberation,” and she praises the body for threading the needle between innovation and public trust (106–9). Yet, she departs from its example in calling for a more truly “public” body that includes lay people in its composition (140). This gets to the more individualistic aspect of Browne’s proposed reform, as the inclusion of lay people in the body is designed to harness the variety of perspectives within a community (152). Browne draws on Arendt and Young in identifying “speculation” regarding “the structural dynamics of everyday legitimate pursuits of private interest” as central to the process (140). Such speculation crucially includes “structural actions” or “those thoughtless actions that operate within the realm of everyday legitimate pursuits of private interest in a given time and place” (25). As in Thiele, narratives or stories play an important role in the process, this time as sources of illumination (141–2).

Neither strictly private nor public, individual nor collective, Koopman focuses on a third domain of reform: technological design, which “affords” or, in other words, encourages or discourages certain forms of use (3, 21). Koopman uses Du Bois’s novel data practices as a model for the conscious attunement of data to the promotion of democratic equality, observing, “Du Bois’s insight—truly striking, given his historical context—is that pursuits of equality making use of data need to also consider the formation of the very data apparatus enrolled in such arguments” (59). This means taking it upon oneself “to collect, collate, and curate data” with the goal of equality in mind (59). Drawing on recent research, Koopman highlights the promise of alternative data formats that promote collaboration by, for example, including collective concepts like ‘group cohesion’ (207). Such formats can promote democratic equality by encouraging individuals to relate to one another as equals.

All of the works here discussed represent important—and hopefully influential— efforts to break the spell of AI’s fairy tale wish. All of the authors marshal extensive empirical support for their arguments and draw on an impressive variety of thinkers while identifying distinct issues of concern and possible responses. Yet, I cannot shake the feeling that, on their own, none of them may be enough. To be fair, on this point, all of the authors may agree. I am not personally afflicted with tech fatalism, so it is not simply because I think AI’s advance is irresistible. Nor is it because the authors desire some kind of quixotic return to a pre-AI world. They do not. The issue, in my opinion, is related to their conceptualization of technology.

At times, all of the authors risk framing recent technological developments, especially AI, as something exterior and alien—a kind of “attack of the blob”—that has appeared out of nowhere to take control of the future from humanity and from which humanity must regain control.Footnote ⁶ But technologies, of which AI is a particularly illuminating example, are not alien invasions. Inextricably intertwined with humanity, as others have argued, AI is better understood not as “artificial” but as “collective” intelligence.Footnote ⁷ It is a social and historical product to which we continue to contribute in our own micro ways with our ceaseless streams of data. Thiele comes close to acknowledging this point in citing the concept of the “distributed brain” (124), and Koopman (the least guilty of this) affirms the collaborative nature of learning (16), but neither grapples with the implications of these ideas, including the fact that the very notion of proprietary corporate ownership of AI may entail its own fairy tale, which elides AI’s inherently collective quality. We need restraints in order to avoid the dangerous temptations of technology, but the success of reforms like “digital distancing,” regulation, and technological redesign is likely to be limited by the current ownership structure. Public ownership of technologies like AI is admittedly—like the solutions proposed in the three books—a tall task for people who seem to have lost their capacity for action, but it may also be a precondition for the success of the three proposals. At worst, the reimagination of technologies like AI as public resources may be a rival fairy tale.Footnote ⁸