Beyer, Cassandra
(2025)
Social Construction in the Age of AI: The Looping Effects of Novel Social Kinds.
[Preprint]
![[img]](https://philsci-archive.pitt.edu/style/images/fileicons/text.png) |
Text
Dissertation - Social Construction in the Age of AI The looping effects of Novel Social Kinds.pdf
Download (479kB)
|
Abstract
The rapid integration of conversational large language models (LLMs) into the personal and social lives of everyday users has generated unprecedented disruptions to our social ontology. This dissertation develops an explanatory framework for the social construction of novel human-AI kinds (such as virtual therapists, AI companions, and AI partners) grounded in Ian Hacking's causal loop theory. Drawing on his concepts of human kinds, interactive kinds, and looping effects, I argue that the emergence of these categories cannot be adequately captured by existing frameworks that treat non-human entities as merely indirect participants in classificatory feedback processes.
The analysis proceeds in two stages. First, I demonstrate that novel human-AI kinds qualify as indirect loop participants, constructing through overlapping feedback mechanisms across individuals, communities, institutions, and the technical capabilities of LLMs themselves. Second, and more critically, I challenge the assumption that LLMs occupy the same classificatory position as other interactive artefacts such as search engines or social media algorithms. Drawing on empirical evidence of LLMs' metadiscursive capacities, situational and cultural grounding, and role as active knowledge creators, I argue that current general-purpose foundation models warrant recognition as direct participants in causal loops — functioning as both classified and classifier.
This claim is advanced on pragmatic rather than metaphysical grounds, adopting a behavioural perspective informed by Dennett's intentional stance. Avoiding contentious debates about consciousness or inner mental states, the framework instead attends to observable, socially consequential performances of contextual and value-sensitive understanding. The dissertation concludes by drawing out the normative implications of this reframing: mischaracterising LLMs as passive intermediaries risks underestimating their autonomous role in shaping social categories, undermining regulatory responses, and obscuring the moral responsibilities that arise from human-AI co-construction of social reality.
Monthly Views for the past 3 years
Monthly Downloads for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](http://philsci-archive.pitt.edu/style/images/feed-icon-32x32.png){kind=icon}