Deconstruct pieces by other writers to evaluate how they were researched, structured, and written
- Ask a chatbot that is powered by a large language model (LLM) like ChatGPT, Claude, Bard, etc., to generate text in the style of a famous writer. It works best to choose a writer whose style is widely known and analyzed, such as Virginia Woolf, Cormac McCarthy, or David Foster Wallace.
- Develop 5 hypotheses (can be several sentences) for why the LLM made certain style emulations. With your hypotheses, be precise and pinpoint particularities of the style (sentence formation, word selection, punctuation and grammar, etc.). Your goal is to deeply consider the work of the author to find parallels with the generated text.
- Read existing style analyses of the authors, for example Virginia Woolf (Columbia’s analysis), Cormac McCarthy (The Conversation), and David Foster Wallace (The New York Times). Write a 500 word analysis comparing your hypotheses to the analyses. Was the LLM effective at emulating the work of the writer? Where did it get things wrong? What important elements of the author’s style were missing in the LLM-generated text?
Optional modifications: Have students write in different distinct styles and prompt the LLM to continue in each style, then analyze the efficacy of the generated writing. See “Reimagining Writing Styles with AI” (slide 93) for more style prompts. Or, ask students to write what they think the LLM might produce, then ask it to do so, and compare.