Digital Defoe: Studies in Defoe & His Contemporaries (Jan 2023)
Defoe and the Chatbot: The Emotional Avoidance of Predictive Prose
Abstract
This article explores the encounter between AI large language models, like ChatGPT, and fiction, which is a massive large language model developed over centuries, across cultures, and with intertextual and contextual references that reach across time, geography, and genre. Both are hypothetical frameworks that rely upon predictive prose, or the “what if.” Both the algorithm and the author imagine what would come next given the situation and the information available. Fictional depictions of artificial intelligence, automata, and chatbots, some orated or published long before the technologies were possible, have shaped our understanding of human-AI interaction, and current AI-human interaction, in turn, is simulated in part based on AI’s understanding of human dialogue as represented in fictional texts it mines for data. Fiction and the literary language of dialogue, then, is influential in how AI communicates. Testing AI’s ability to recognize and analyze fiction brings to light the complexity of literature. Daniel Defoe’s prose and use of the subjunctive mood in moments of dialogue provides a revealing test case for the limitations of AI analytical abilities. Defoe often relies upon hypothetical constructions, like mandative subjunctives (“I wish that”), modal auxiliaries (“would,” “could”), and conditionals (“if this then that”) when characters are in emotional situations. Inspired by the chatbot-user dialogue that takes place in ChatGPT, and its struggle to articulate the meaning of key literary scenes in which characters shift into the subjunctive mood, this article finds that Defoe’s use of subjunctive constructions interrupts the emotional connection of the speakers, preventing them from reaching empathetic understanding of the other. The hypothetical, then, in literary dialogue and also in AI-human “chat,” creates emotional disruption and resistance to empathy. The article concludes by questioning whether AI’s struggles with fiction may lead to other realizations about the sophistication of literary language and narrative.
