Research / Creativity
Creativity
What can AI do as a creative agent — and what can it not do? The writer's Turing test, the AI Fiction Paradox, and the public conversations on AI and creativity at UNESCO, Al Jazeera, and elsewhere.
The writer's Turing test
- “Can GPT-3 Pass a Writer's Turing Test?” (Elkins and Chun, Journal of Cultural Analytics 5.2, September 2020) — appeared four months after the release of GPT-3. Proposed the writer's Turing test as a framework for evaluating large language model creative competence; provides the first systematic empirical evaluation of GPT-3 in humanities-relevant creative writing tasks. Cited by Luciano Floridi and Massimo Chiriatti in Minds and Machines 30 (2020); by Terence Tao and Tanya Klowden at the conceptual pivot of their essay for the Blackwell Companion to the Philosophy of Mathematics (2026); by Martin Paul Eve at footnote 88 of The Digital Humanities and Literary Studies (Oxford University Press, 2022). The writer's Turing test concept has been adopted as a named framework across philosophy of AI and NLP evaluation.
The AI Fiction Paradox
- “AI Fiction Paradox” (Elkins, arXiv 2603.13545, 2026) — names what AI-generated fiction systematically cannot do, identifying failures in narrative causation, informational revaluation, and multi-scale emotional architecture. Cross-listed with Storytelling.
What LLMs do to authorship
- “AI Comes for the Author” (cross-listed with Mind) — authorship as a creative and cognitive category under conditions of LLM generation.
Theatrical AI
- DIVAbot (Arts at Denison, January 25, 2021) — one of the earliest staged uses of transformer-based generative AI in theatrical performance in the United States, twenty-two months before ChatGPT.
Public engagement on AI and creativity
- UNESCO appearances on AI and creative production (also Governance)
- Al Jazeera segment on AI and creativity
- Public talks on AI and the arts at numerous venues