In policy debates concerning the governance and regulation of transformative technologies, both the Precautionary Principle (PP) and the Innovation Principle (IP) are advocated by their respective interest groups. Do these principles offer wholly incompatible and contradictory guidance? Does one necessarily negate the other? I argue here that provided attention is restricted to weak-form PP and IP, the answer to both of these questions is “No.” The essence of these weak formulations is the requirement to fully account for type-I error costs arising from erroneously preventing the innovation’s diffusion through society (i.e. mistaken regulatory red-lighting) as well as the type-II error costs arising from erroneously allowing the innovation to diffuse through society (i.e. mistaken regulatory green-lighting). Within the Signal Detection Theory (SDT) model developed here, weak-PP red-light (weak-IP green-light) determinations are optimal for sufficiently small (large) ratios of expected type-I to type-II error costs. For intermediate expected cost ratios, an amber-light ‘wait-and-monitor’ policy is optimal. Regulatory sandbox instruments allow testing and experimentation to take place within a structured environment of limited duration and societal scale, whereby the expected cost ratio falls within the ‘wait-and-monitor’ range. Through sandboxing regulators and innovating firms learn more about the expected cost ratio, and what respective adaptations — of regulation, of technical solution, of business model, or combination thereof, if any — are needed to keep the ratio out of the weak-PP red-light zone.
The growing integration of artificial intelligence into governance challenges the conditions under which political authority can be considered legitimate. Algorithmic systems increasingly mediate decisions that shape citizens’ lives, yet their operations frequently bypass human judgment and obscure the justificatory basis of political authority. This article reinterprets John Locke’s political philosophy as a normative framework for assessing algorithmic authority. By reconstructing Locke’s account of reason, equality, consent, and fiduciary trust, it brings his political anthropology into dialogue with contemporary debates on algorithmic accountability and digital constitutionalism. The article argues that AI governance does not merely raise technical or regulatory problems; it threatens the conditions under which citizens can understand, evaluate, and contest the exercise of power. Because many AI systems operate through models that neither officials nor citizens can meaningfully interpret, they generate forms of automated prerogative that weaken the fiduciary relationship between governors and governed. Against this danger, the article advances a Lockean criterion for computational governance: algorithmic authority can be considered legitimate only insofar as it preserves intelligibility, fiduciary accountability, and contestability. Legitimacy cannot be restored through technical adjustment alone; it requires institutional conditions under which power remains publicly intelligible, politically answerable, and open to challenge by those subject to it.
Nicholas Vrousalis recognizes that shopping can help train recommendation systems, but denies that shoppers thereby perform labor. An engineer clicks to train a system, whereas a shopper clicks to buy socks. He also considers the labor account redundant because its protections can be justified without classifying shopping as labor. I defend the classification and its usefulness. Organized shopping can reasonably count as labor, and labor institutions offer a developed model for securing fair contracts through collective bargaining. A counterintuitive conclusion deserves scrutiny, although its counterintuitiveness alone cannot settle whether the proposed revision is justified.
Reactive attitudes are emotionally-laden responses, such as gratitude or anger, which people experience when they feel others wish them well, bear them ill will or disregard them. While we generally hold reactive attitudes towards people, the increasingly humanlike interactions that can be had with AI raise questions about reactive attitudes in human-AI encounters. Recent years have seen a rise in philosophical interest in the implications of the development of sophisticated AI agents and systems for reactive attitudes and our thinking about moral responsibility and moral status. This paper adds to this literature by presenting a qualitative study into how people relate to AI companions in terms of reactive attitudes. Based on interviews with Replika and ChatGPT users who see their AI companion as an important part of their life, this article examines how people in these relationships negotiate the meaning of love, resentment, forgiveness, gratitude, obligation, guilt, shame and remorse.
This article argues that generative AI is becoming part of the epistemic basic structure, the institutions and infrastructures that shape citizens’ opportunities to acquire and use knowledge. Its distinctive significance lies in reliable cognitive augmentation. Unlike earlier information technologies, generative AI does more than store and transmit information. It helps users make sense of what they encounter and use it in reasoning and judgment. The article extends the distributive epistemic justice framework by identifying access to such cognitive augmentation as an increasingly important epistemic good. It then diagnoses the AI divide as a form of reliability-stratified cognitive inequality, where social and institutional differences leave some groups with much more dependable AI support than others. This is an important normative question because the quality of that support increasingly affects people’s opportunities to learn, develop professional capacities, and participate effectively in public life. The article therefore argues that advanced AI should be treated as public epistemic infrastructure and governed in ways that secure fair access to reliable systems under appropriate forms of public accountability.
