Lorenzo Manuali

An increasingly large number of projects seek to use Large Language Models (LLMs) to enhance or support democracy. I argue that many of the ways in which computer scientists, deliberative democratic practitioners, and others are using LLMs to do this threatens the non-instrumental value of a collective’s ability to determine its own future. In particular, I present a novel worry that projects aimed at algorithmically facilitating deliberation and representing people’s interests in political processes threaten what has been called democratic autonomy. I begin with some conceptual groundwork concerning collective self-determination (and specifically democratic autonomy) to motivate its non-instrumental value. Next, I offer a few necessary conditions for democratic autonomy from the literature, such as the possession of a joint intention and said joint intention being realized in policy at least some significant portion of the time. I then show when two kinds of projects that use LLMs to enhance democracy – facilitative and representative LLMs – threaten these necessary conditions and thus democratic autonomy. I conclude by outlining some practical upshots and recommendations for projects that aim to use LLMs to enhance or support democracy.

LLMs are among the most advanced tools ever devised for understanding and generating natural language. Democratic deliberation and decision-making involve, at several distinct stages, the production and comprehension of language. So it is natural to ask whether our best linguistic tools might prove instrumental to one of our most important linguistic tasks involving language. Researchers and practitioners have recently asked whether LLMs can support democratic deliberation by leveraging abilities to summarise content, to aggregate opinions over summarised content, and to represent voters by predicting their preferences over unseen choices. In this paper, we assess whether using LLMs to perform these and related functions really advances the democratic values behind these experiments. We suggest that the record is mixed. In the presence of background inequality of power and resources, as well as deep moral and political disagreement, we should not use LLMs to automate non-instrumentally valuable components of the democratic process, nor should we be tempted to supplant fair and transparent decision-making procedures that are practically necessary to reconcile competing interests and values. However, while LLMs should be kept well clear of formal democratic decision-making processes, we think they can instead strengthen the informal public sphere—the arena that mediates between democratic governments and the polities that they serve, in which political communities seek information, form civic publics, and hold their leaders to account.

In this paper, I propose an account of behavioral addiction in terms of what I call addictive motivational scaffolds (AMSs). Taking inspiration from recent work concerning psychiatric externalism and addiction, I propose and describe the concept of motivational scaffolding: external structure that enhances, supports, or regulates motivational processes in the mind-brain. I then argue that some motivational scaffolds are likely difference-makers in that they make an activity more addictive. The paper proceeds in three main parts. First, I describe the concept of a motivational scaffold and how it builds on recent literature in 4E cognition/psychiatric externalist accounts of addiction. Using gambling and gaming as paradigm cases of addictive activities, I then identify and empirically justify four addictive motivational scaffolds (AMSs): (1) quantified metrics, (2) reward uncertainty, (3) short time-horizon to reward, and (4) physically salient features. Finally, I apply my account to social media to showcase its philosophical usefulness: analyzing behavioral addiction in terms of AMSs uniquely elucidates the more _structural_ aspects of the addictiveness of social media, which are undertheorized.

ABSTRACT This paper uses abduction to propose a theory of addictive desire that is based on a dual dysfunction in learning and motivation. The theory is termed the Coherent Dual Theory of Addictive Desire (CDT) because it proposes that a dual dysfunction lies at the heart of addictive desire. This paper first lays out CDT in psychological terms. It then focuses on how CDT coheres with both the neurobiology and phenomenology of addictive desire, granting the theory empirical support and more explanatory power. The paper concludes by addressing an objection from opponent process theory. The result is a theory of addictive desire that contains multiple abductive virtues: It is simple, general, testable, and coherent on the levels of neurobiology, psychology, and phenomenology.