Annette Zimmermann

Democratizing AI offers a powerful rethinking of how artificial intelligence should be governed. Challenging the dominance of tech elites in shaping AI's future, Zimmermann argues that AI deployment is a political act-one that must be subject to democratic control. They propose a practical "playbook" for reclaiming agenda-setting power through civic participation, public ownership, and institutional reform. Engaging with leading critics, Zimmermann defends a risk-sensitive proceduralist approach while acknowledging the deeper structural challenges posed by capitalism, inequality, and democratic fatigue. This book is a call to action: to resist learned helplessness, confront techno-authoritarianism, and shape AI's trajectory in line with democratic values. Thoughtful, urgent, and visionary, the book is essential reading for anyone concerned with the future of technology and democracy.

The claim that the AI community, or society at large, should ‘democratize AI’ has attracted considerable critical attention and controversy. Two core problems have arisen and remain unsolved: conceptual disagreement persists about what democratizing AI means; normative disagreement persists over whether democratizing AI is ethically and politically desirable. We identify eight common AI democratization traps: democratization-skeptical arguments that seem plausible at first glance, but turn out to be misconceptions. We develop arguments about how to resist each trap. We conclude that, while AI democratization may well have drawbacks, we should be cautious about dismissing AI democratization prematurely and for the wrong reasons. We offer a constructive roadmap for developing alternative conceptual and normative approaches to democratizing AI that successfully avoid the traps.

It is becoming more common that the decision-makers in private and public institutions are predictive algorithmic systems, not humans. This article argues that relying on algorithmic systems is procedurally unjust in contexts involving background conditions of structural injustice. Under such nonideal conditions, algorithmic systems, if left to their own devices, cannot meet a necessary condition of procedural justice, because they fail to provide a sufficiently nuanced model of which cases count as relevantly similar. Resolving this problem requires deliberative capacities uniquely available to human agents. After exploring the limitations of existing formal algorithmic fairness strategies, the article argues that procedural justice requires that human agents relying wholly or in part on algorithmic systems proceed with caution: by avoiding doxastic negligence about algorithmic outputs, by exercising deliberative capacities when making similarity judgments, and by suspending belief and gathering additional information in light of higher-order uncertainty.

The democratic boundary problem raises the question of who has democratic participation rights in a given polity and why. One possible solution to this problem is the all-affected principle, according to which a polity ought to enfranchise all persons whose interests are affected by the polity’s decisions in a morally significant way. While AAP offers a plausible principle of democratic enfranchisement, its supporters have so far not paid sufficient attention to economic participation rights. I argue that if one commits oneself to AAP, one must also commit oneself to the view that political participation rights are not necessarily the only, and not necessarily the best, way to protect morally weighty interests. I also argue that economic participation rights raise important worries about democratic accountability, which is why their exercise must be constrained by a number of moral duties.

Disagreement persists about when, if at all, disenfranchisement is a fitting response to criminal wrongdoing of type X. Positive retributivists endorse a permissive view of fittingness: on this view, disenfranchising a remarkably wide range of morally serious criminal wrongdoers is justified. But defining fittingness in the context of criminal disenfranchisement in such broad terms is implausible, since many crimes sanctioned via disenfranchisement have little to do with democratic participation in the first place: the link between the nature of a criminal act X (the ‘desert basis’) and a fitting sanction Y is insufficiently direct in such cases. I define a new, much narrower account of the kind of criminal wrongdoing which is a more plausible desert basis for disenfranchisement: ‘political wrongdoing’, such as electioneering, corruption, or conspiracy with foreign powers. I conclude that widespread blanket and post-incarceration disenfranchisement policies are overinclusive, because they disenfranchise persons guilty of serious, but non-political, criminal wrongdoing. While such overinclusiveness is objectionable in any context, it is particularly objectionable in circumstances in which it has additional large-scale collateral consequences, for instance by perpetuating existing structures of racial injustice. At the same time, current policies are underinclusive, thus hindering the aim of holding political wrongdoers accountable.

This dissertation develops and defends four interconnected arguments about the All-Affected Principle (AAP): the view that every person whose morally weighty interests are affected by a democratic decision has the right to participate in that decision. The first part (“Narrow Possibilism about Democratic Enfranchisement”) examines how we should distribute democratic participation rights: a plausible version of AAP must avoid treating unlike cases alike, which would be procedurally unfair. The solution is to distribute participation rights proportionately to the risk that a person’s interests will be affected. AAP thus implies an account of political equality that requires adherence to the ‘one person—one vote’ model only if interests are indeed equally affected. The second part (“Economic Participation Rights and the AAP”) argues that AAP supporters have paid insufficient attention to economic participation rights. The exercise of such rights raises unique worries about democratic accountability, which is why their exercise is constrained by a number of duties. The third part (“What AAP Is, and How (Not) to Fight It”) explores how AAP fares in light of possible objections from desirability and feasibility. Unlike crude versions of AAP, a plausibly restricted version of AAP cannot be dismissed as easily as many AAP sceptics may have thought. My reflections here are useful for AAP supporters and sceptics alike: this paper helps clarify what kind of objection can cast serious doubt on AAP. The fourth part (“Criminal Disenfranchisement, Political Wrongdoing, and Affected Interests”) asks: is AAP compatible with criminal disenfranchisement? AAP, when endorsed in combination with a plausible theory of punishment, is compatible with disenfranchising a narrow set of criminal wrongdoers only: those guilty of ‘political wrongdoing’, which is wrong primarily because it undermines democratic procedures and institutions for private gain. The upshot is that current blanket policies of criminal disenfranchisement are incompatible with AAP.