On March 10, Hannes Bajohr, Assistant Professor in the Department of German at UC Berkeley, presented his work on “Artificial and Post-artificial Texts”, considering the reader’s expectations after AI.
“In this seminar, Hannes Bajohr explores how the widespread use of AI-generated texts is transforming the expectations readers have when encountering written language. Traditionally, readers assume that a text is written by a human, with intent, meaning, and a traceable authorship—a belief system Bajohr calls the 'standard expectation.' This assumption begins to unravel as AI systems produce more texts that mimic human writing, prompting a crisis of interpretation and authenticity. Drawing on historical context, he references Max Bense’s concept of artificial poetry and Theo Lutz’s 1959 computer-generated poems to illustrate how early experiments already hinted at today’s dilemmas. Bajohr introduces the notion of 'strong deception' where AI aims to convincingly pass as human, and contrasts it with 'banal deception,' where users knowingly engage with machine text but still suspend disbelief for practical interaction.
As trust in textual origin erodes, readers may adopt a 'post-artificial' stance—one that no longer fixates on who wrote the text but focuses solely on the text’s content. Bajohr ultimately calls for a recalibration of critical reading strategies, emphasizing the need to adapt our interpretive habits without succumbing to technological determinism or the myth of AI neutrality.”
Some of the discussion after Bajohr’s seminar interrogated the difference between non-fiction writing and fiction. Much of the analysis of AI has focused on its accuracy. But for fiction the point of the story is the story. There is no ground truth to assess the model against. Bajohr linked the question to ghost writers today. Do we appreciate their stories because they’re written under a different name? Why? Why not?
The audience also explored with Hannes the ways in which AI can or can not produce narrative. In the technical sense all large language models are correlative and therefore can not create causal relationships, with narrative being a series of causal relationships. However, it has long been observed that it is the reader who produces most of the causality when they read a text. The degree to which it is present in the text is hard to answer because it changes depending on how the reader receives the story.
Watch the video below for all of Bajohr’s considerations of how language is changing in the technical age we inhabit today, and which will continue to unfold “both without fearing technology but also refusing to succumb to some of its ideologies.”
In that context, one thing seems certain to me with the increasing penetration of language technologies, with the triumph of AI models. Our expectations as readers will change. – Prof Hannes Bajohr, human intelligence, received by you, the reader.
