Writing and the Machine

Patricia Hodges

A challenge the reader of an AI-generated (Large Language Model) text faces is what T S Eliot called the ‘dissociation of sensibility’, a term that describes a separation between intellect and emotion. Eliot argued that after the 17th century, English poetry lost its capacity to unite logical thought with sincere, visceral feeling. This thought reminds me that the algorithm possesses computer reasoning and theoretical knowledge; it can efficiently manipulate language and recall facts, but it misses sensory perception, lived experience and emotional investment. The result is a blend of human conceptual thought over time and history, rather than a true reflection of human perception and thought in action. Authentic expression blends intellect with emotion while LLMs’ output is strictly grammatical, precise and structured, an imitation of sentient experience.

Moreover, we traditionally view texts as communication between human beings: we naturally look for a specific sender and an intended audience, a concept that helps define authorship in
Western culture. With AI-generated content, this conventional model breaks up: the sender becomes an algorithmic combination of a myriad human voices. The point of view often changes: sometimes it indicates the majority consensus of the data, and at other times it mimics the specific persona requested by the user. The user becomes the ‘ideal reader’, a literary theoretical concept that refers to an imaginary, perfect recipient who has all the cultural knowledge and emotional intelligence to be able to decode a text precisely as the writer intended. But unlike a human author who has a consistent natural voice, the LLM text can completely distort its ideal reader based on the user’s instructions and prompts. For example, asking an LLM to write a peer-reviewed style piece changes the ideal reader into an expert who can deal with complex syntax and specialist terminology (and wants no bullet points). At the other extreme, prompting it to write an engaging little post, changes the ideal reader into the proverbial phone-scrolling lover of ultra-short sentences and strong emotive hooks.

An attractive and efficient shortcut, yes: using AI technology for writing can significantly reduce the hassle of drafting, editing, formatting and proofreading, and can even help to break through a blank page, the writer’s nightmare. LLMs certainly present a new way to do and to think about language, but as Keats stated in 1817: “What quality went to form a Man of Achievement, especially in Literature […] was the capability of being in uncertainties, mysteries, doubts, without any irritable reaching after fact and reason.” In this view, ‘negative capability’ is essential for truly creative writing rather than a simulation driven by statistical probability.