Conference Agenda

As a co-product of the interplay between knowledge and power, algorithmic governance raises fundamental questions of political epistemology while offering technical solutions constrained by value norms. Political epistemology, as an emerging interdisciplinary field, investigates the possibility of political cognition by addressing issues such as political disagreement, consensus, ignorance, emotion, irrationality, democracy, expertise, and trust. Central to this inquiry are the political dimensions of algorithmic governance and how it shapes or even determines stakeholders’ political perceptions and actions. In the post-truth era, social scientists have increasingly employed empirical tools to quantitatively represent algorithmic political bias and rhetoric.

Despite advancements in the philosophy of technology, which has shifted from grand critiques to micro-empirical studies, it has yet to fully open the space for political epistemological exploration of algorithmic governance. To address this gap, this paper introduces power topology analysis. Topology, a mathematical field that studies the properties of spatial forms that remain unchanged under continuous transformation, has been adapted by thinkers like Gilles Deleuze, Michel Foucault, Henri Lefebvre, Bruno Latour, and David Harvey to examine the isomorphism and fluidity of power and space. Power, like topology, retains continuity even through transformations, linking the two conceptually.

This paper is structured into four parts. The first explores the necessity and significance of power topology in conceptualizing algorithmic power and politics through the lens of political epistemology. The second examines the generative logic and cognitive structure of power topology within algorithmic governance. The third analyzes how power topology transforms algorithmic power relations into an algorithmic political order. The fourth proposes strategies for democratizing algorithmic governance through power topology analysis.

The introduction of power topology analysis offers a reflexive perspective for the philosophy of technology to re-engage with political epistemology—an area insufficiently addressed by current quantitative research and ethical frameworks. This topological approach provides a detailed portrait of algorithmic politics by revealing its power topology. Moreover, it redefines stakeholder participation by demonstrating how algorithms stretch, fold, or distort power relations, reshaping the political landscape. By uncovering the material politics of these transformations, power topology encourages the philosophy of technology to reopen political epistemological spaces and adopt new cognitive tools for outlining the politics of algorithmic governance. Ultimately, this framework aims to foster continuous, rational, and democratic engagement by stakeholders in the technological transformation of society, offering a dynamic and reflexive tool for understanding the intersection of power, politics, and algorithms.

Recent advancements in AI systems, in particular Large Language Models, have sparked renewed interest in a technological vision once confined to science fiction: generative AI agents capable of simulating human personalities. These agents are increasingly touted as tools with diverse applications, such as facilitating interview studies (O’Donnell, 2024), improving online dating experiences (Batt, 2024), or even serving as personalized "companion clones" of social media influencers (Writer, 2023). Proponents argue that such agents, designed to act as "believable proxies of human behavior“ (Park et al. 2023) offer unparalleled opportunities to prototype social systems and test theories. As Park et al. (2024) suggest, they could significantly advance policymaking and social science by enabling large-scale simulation of social dynamics.

This paper critically examines the foundational assumptions underpinning these claims, focusing on the concept of believability driving this research. What, precisely, does "believable" mean in the context of generative agents, and how might an uncritical acceptance of their believability create self-fulfilling prophecies in social science research? This analysis begins by tracing the origins of Park et al.’s framework of believability to the work of Bates (1994), whose exploration of believable characters has profoundly influenced the field.

Drawing on Günther Anders’ (1956) critique of technological mediation and Donna Haraway’s (2018, 127) reflections on "technoscientific world-building“, this paper situates generative agents as key sites where science, technology, and society intersect. Ultimately, it calls for a critical reexamination of the promises and perils of generative agents, emphasizing the need for reflexivity in their conceptualization, as well as their design and application. By interrogating the assumptions behind believability, this research contributes to a deeper understanding of the socio-technical implications of these emerging AI systems.

Building on Louise Amoore’s (2020) concept of algorithms as composite creatures, this paper explores the implications of framing generative agents as "believable." In the long run, deploying these AI systems in social science research risks embedding prior normative assumptions into empirical findings. Such feedback loops can reinforce preexisting models of the world, presenting them as objective realities rather than as socially constructed artifacts. The analysis highlights the danger of generative agents reproducing and amplifying simplified or biased representations of complex social systems, thereby shaping policy and theory in ways that may perpetuate these distortions.

References

Amoore, Louise (2020). Cloud Ethics: Algorithms and the Attributes of Ourselves and Others. Durham: Duke University Press.

Anders, Günther (1956). Die Antiquiertheit des Menschen Bd. I. Munich: C.H. Beck.

Batt, Simon (2024). „Bumble Wants to Send Your AI Clone on Dates with Other People's Chatbots.” Retrieved from https://www.xda-developers.com/bumble-ai-clone-dates-other-peoples-chatbots/.

Contreras, Brian (2023). „Thousands Chatted with This AI ‘Virtual Girlfriend.’ Then Things Got Even Weirder.” Retrieved from https://www.latimes.com/entertainment-arts/business/story/2023-06-27/influencers-ai-chat-caryn-marjorie.

Haraway, Donna Jeanne (2018). Modest_Witness@Second_Millennium. FemaleMan_Meets_OncoMouse: Feminism and Technoscience. Second edition. New York, NY: Routledge, Taylor & Francis Group.

O’Donnell, James (2024). „AI Can Now Create a Replica of Your Personality.” Retrieved from https://www.technologyreview.com/2024/11/20/1107100/ai-can-now-create-a-replica-of-your-personality/.

Park, Joon Sung, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein (2023). „Generative Agents: Interactive Simulacra of Human Behavior.“ In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology, 1–22. https://doi.org/10.1145/3586183.3606763.

Park, Joon Sung, Carolyn Q Zou, Aaron Shaw, Benjamin Mako Hill, Carrie Cai, Meredith Ringel Morris, Robb Willer, Percy Liang, and Michael S Bernstein (2024). „Generative Agent Simulations of 1,000 People“, Retrieved from arXiv. https://doi.org/10.48550/arXiv.2411.10109.