By now, pointing out the ubiquity and sophistication of algorithmic decision-making—and the challenges it presents—seems obvious. Alongside public-facing experiments like ChatGPT and DALL-E, there is no shortage of excellent work by scholars and cultural critics attempting to unpack the promises, pitfalls, and possibilities of AI technology. In Algorithmic Reason: The New Government of Self and Other, Claudia Aradau and Tobias Blanke contribute to these ongoing discussions by offering their own conceptual framework—the titular “algorithmic reason”—with which to make sense of the ways that algorithms are transforming not just the way we write term papers or make art, but the very way we think about the world.
While much of the public discussion around algorithms and public life focuses on the proper role and scope of algorithmic decision-making, as well as “how to ensure its legitimacy, efficiency, and fairness” (p. 9), Aradau and Blanke instead approach algorithms as embodying a particular political rationality. “Algorithmic reason,” they argue, is a “new government of self and other” (p. 3) that “redraws the boundaries among those to be brought within the remit of government: the part and the whole, the individual and the population, self and other” (p. 8). Just like other political rationalities—like neoliberalism and statistical reasoning—algorithmic reason is a way of knowing that has implications for how we govern and are governed by others.
But the novelty of algorithmic reason, argue Aradau and Blanke, is that it blurs the distinction between large and small. Instead of focusing exclusively, or even primarily, on the large-scale population or the single individual, algorithms move continuously between the two scales: they both break down, or “decompose,” the large into the small and “recompose,” or reassemble, the small into the large. Rather than remaining limited to either the macro- or micro-level, algorithmic reason dismantles the very distinction between these two categories. As a “political rationality,” then, algorithmic reason “produces knowledge about individuals and populations to conduct their conduct and enables decisions that draw lines between self and other” (p. 14). The result, Aradau and Blanke show throughout the book, is a breakdown of conventional political categories into more amorphous and ever-shifting distinctions: enemies become “anomalies,” infrastructure becomes “platforms,” and “value” is generated not from new content but merely new combinations.
To make their case, Aradau and Blanke organize the book around a series of “scenes” through which they demonstrate, first, how algorithmic reason is shifting governance techniques in, for example, predictive policing (Chapter 2), national security (Chapter 3), and the platform economy (Chapters 4 and 5). In the final section of the book, the authors suggest ways that current approaches to addressing certain technological challenges fail to account for the effects of algorithmic rationality. Instead, through discussions of AI ethics (Chapter 6), facial recognition accountability (Chapter 7), and state and platform governance practices (Chapter 8), Aradau and Blanke offer strategies of “friction, refusal, and resistance” that work to counter algorithmic reason by “enable[ing] the political formation of algorithms as public things” (pp. 217-218).
Aradau and Blanke’s insistence on the democratic potential of algorithms, despite their critiques, is one of Algorithmic Reason’s most welcome interventions. Too often, critics of AI and of algorithms more generally emphasize the ways that they undermine democratic politics. While these critiques are admittedly compelling, Aradau and Blanke provide a conceptual framework that can help articulate these challenges while still leaving room for political action to address them. Algorithmic Reason suggests that we consider algorithms not as something imposed on us by experts who ultimately retain responsibility (and control), but rather as “public things” that gather groups of people into specific kinds of political relationships.
Attending to the problems with algorithms, in this reframing, is thus not (only) a problem of identifying the right set of ethical principles to, say, govern AI decision-making. Rather, it is (also) one of fostering the kinds of spaces that make visible the publics around algorithms, rendering the algorithms objects of contestation and collective action by a plurality of constituencies. One model for this is Aradau and Blanke’s example of a “hackathon” (Chapter 6) that brought together users and technologists to discuss the technical operations of algorithms and their effects “in a collective setting” (p. 157). Such interventions serve as productive strategies to address algorithmic harms because they specifically account for the political dimensions involved—the collective disagreement, deliberation, and action that are part and parcel of democratic life.
Examples such as the hackathon are present throughout Algorithmic Reason, due to Aradau and Blanke’s novel methodological approach. Drawing from Rancière’s concept of the “scene,” the authors use a series of rich and provocative examples to “trace how algorithmic variations inflect and hold together heterogenous practices of governing across time and space” (p. 14). From the Cambridge Analytica scandal (Chapter 1) to Spotify patents (Chapter 5) to Google employees’ petitions (Chapter 6) and class-action lawsuits by Facebook content moderators (Chapter 8), Algorithmic Reason takes up a range of debates that have captured the public’s imagination regarding algorithms. Through these scenes, Aradau and Blanke explore the tensions at work in disparate practices of algorithmic reason. In so doing, they reframe the debates about algorithms away from what they are doing—the inputs, outputs, and calculations—and focus instead on how algorithms are doing it, including the way that logic may (re)shape our political vision.
But while Algorithmic Reason’s political orientation and provocative examples help to underscore the value of Aradau and Blanke’s framework for rethinking the role of algorithms in public life, the practical implications of this work remain underdeveloped in the text. This is, in part, a question of method: Aradau and Blanke are clear that their use of the “methodology of the scene” is intended to help “attend to how controversies unfold” (p. 208) in ways that reveal “a trajectory of algorithmic reason as undetermined” (p. 218). Algorithmic Reason, then, works to give readers a new vocabulary with which to understand algorithms and their effects. But it remains to be seen what this means for practically addressing the challenges that Aradau and Blanke identify. How does attending to algorithmic reason as a system of governance change the way we approach AI governance? How do we scale up spaces like the hackathon to meet the scope of mass society on which algorithms operate? And what is the role of experts in this work?
That Algorithmic Reason generates such questions testifies to the richness of the framework that Aradau and Blanke provide. And, unlike many who study algorithms, they conclude on an optimistic note: despite its influence, “algorithmic reason does not undo democracy, reflexivity, or political action” (p. 218). Rather, by attending to the effects of algorithmic reason—the way it radically shifts the categorization we use to understand the world—we can counteract some of its tendencies. The goal, then, is not to find a single set of principles through which to make algorithms “legitimate,” but rather to create spaces within which we can treat algorithms as the public things they are—to ensure that we properly understand them as objects of collective concern.
