Logo image
Back
Fragile texts and machine readers: Trans/in/dividual reading tactics in a complex technical milieu
Book chapter

Fragile texts and machine readers: Trans/in/dividual reading tactics in a complex technical milieu

Elizabeth de Freitas, Matthew X. Curinga, Maggie MacLure, David Rousell, Laura Trafí-Prats and Sarah E. Truman
Posthuman Social Science and Computational Culture: Essays on Methodology, Theory and Practice, pp.212-224
Routledge, 1st
2026
Appears in  Recent Faculty of Education Publications

Metrics

1 Record Views

Abstract

There is little doubt that generative neural network algorithms like large language models (LLM) are good at predictive aspects of language processing. 1 They scrape information from diverse digital registers and achieve a synthesis akin to reading and writing (Pavlick, 2023). The LLM has emerged as part of our investment in neural network approaches to AI, enhanced by transformer algorithms that incrementally map the syntax of its near-infinite digital resources. As part of their training, these algorithms scour the internet, sweep through texts of dubious authority, and rummage through our language rubble. Generative algorithmic infrastructure allows them to simulate and mimic all that they have read. It may sound glib, but I fear that LLMs are the readers who are best suited to our distraction economy and post-truth conditions. Ironically, they operationalize an algorithmic architecture called an attention mechanism that decomposes grammar and lexicon into a scheme of plausible “relevance” measures (Vaswani et al., 2017). These “transformers” embody a kind of neoliberal relativism in their performative attention to variety and viral influence, and their affirmation of internet alternative truths. Some advocates argue that LLMs have achieved an “implicit embodied knowledge” of language, and that they are very effective at simulating a “grounded” reading, but there are all kinds of debates about how they do so and the extent to which their language use can be characterized by traditional linguistic and cognitive theories (Millière, 2023). Critics of LLMs suggest that neural network models will always produce poor examples of AI (Chollet, 2019; Marcus, 2022), being stochastic parrots and inefficient learners, lacking symbolic and logical structure as well as any grounded semantics (Bender et al., 2021). And yet our understanding of how these technical models achieve their language skills is nascent, and it feels far too premature to dismiss their cognitive behavior as utterly shallow.

Details

Logo image