Political Theory

H01(c) - Ethics and Representation in a Digital Era

Date: Jun 3 | Time: 08:30am to 10:00am | Location:

Chair/Président/Présidente : Caleb J. Basnett (Mount Allison University)

Discussant/Commentateur/Commentatrice : Joshua Ayer (University of Alberta)

From the Physical World to the Infosphere: A Reevaluation of Hubert Dreyfus's Critique of AI in the world of Large Language Models: Greg Dinsmore (Université de l'Ontario français)
Abstract: The last decade has witnessed remarkable advancements in machine learning techniques, particularly in deep learning. Large language models (LLMs), such as ChatGPT and Claude, have rekindled debates about the nature of intelligence and the potential for computers to develop artificial general intelligence (AGI). In a series of books and articles beginning with Alchemy and AI in 1965, philosopher Hubert Dreyfus critiqued the emerging AI research program for its limited understanding of human intelligence. Drawing on the ideas of Martin Heidegger and Maurice Merleau-Ponty, Dreyfus argued that AI research misunderstood the fundamentally situated and embodied nature of human intelligence. This paper aims to reexamine Dreyfus’s critique of AI to adapt it to contemporary models of artificial intelligence and the issues posed by an increasingly digitalized world. If human intelligence is conditioned by our embodiment in a human world, it may be possible to understand it as situated within a digital world (an "infosphere" in the sense of Luciano Floridi). Large language models are better understood not as a step toward the development of AGI but as a tool for navigating a world increasingly shaped by digital technologies.

Digital Twins: Can Robots Represent Us? : Michael MacKenzie (Vancouver Island University), Jasmine Tomczyk (Vancouver Island University)
Abstract: In César Hidalgo’s (2018) theory of “augmented democracy” individuals would train their own digital twins to act for them in the political sphere. These digital twins would express, represent and predict each person’s unique constellation of political views and make or influence political decisions for them. Hidalgo argues that digital twins would help solve the cognitive bandwidth problem, thus making it possible for more people to be involved in more political decisions through their digital representatives. Recent developments in Artificial Intelligence (AI) have made this, and other models of digital representation technically feasible. In this paper we explore whether digital democratic twins are desirable on democratic grounds. We sidestep difficult questions about ownership and security to focus on normative questions in democratic theory. Digital twins would make personalized representation possible and potentially enhance political deliberations. Digital twins would not be marred by the (same) cognitive and social shortcomings that affect and often derail deliberations between humans. Despite these potential advantages, we argue that digital twins cannot meet the normative demands of good democratic representation. Meaningful and legitimate representation requires some basis of shared experience between the represented and their representatives. Furthermore, representatives must be capable of taking on moral obligations to — and thus responsibilities towards — those they represent. If AI agents were to meet these criteria in the future they would have a case to be represented in our political systems, which would undermine the possibilities of personalized representation that underpin the argument for digital twins in the first place.