Published February 22, 2026 | Version preprint-v1
Preprint Open

Orphaned Sophistication: Detecting AI-Generated Prose Through Structurally Unsupported Figurative Language

Authors/Creators

  • 1. Independent Researcher

Description

We introduce orphaned sophistication, a novel linguistic feature for distinguishing AI-generated prose from human-authored text. Orphaned sophistication occurs when a figurative expression (metaphor, personification, synecdoche) appears without the network of supporting choices — lexical, syntactic, tonal — that would normally license it in skilled human writing. We operationalise the feature through a three-component annotation scheme (Structural Integration, Tonal Licensing, Lexical Ecosystem) and present a hand-annotated corpus of 400 passages drawn equally from literary fiction and LLM outputs across four model families. Inter-annotator agreement is high (Cohen's kappa = 0.81). A logistic-regression classifier using only orphaned-sophistication scores achieves 78.2% balanced accuracy, and when combined with existing stylometric baselines the feature yields a statistically significant 4.3 percentage-point improvement (p < 0.01). Qualitative analysis confirms that the feature captures a failure mode — figurative language produced without underlying rhetorical commitment — that surface-level detectors miss. We release the annotation guidelines, corpus, and scoring code to support further work on interpretable AI-text detection.

Notes

Preprint. Under review.

Files

Quinn_Orphaned_Sophistication_2026_preprint.pdf

Files (206.2 kB)

Name Size
md5:90dc07417614c286719d98bf6ab7cee5
206.2 kB Preview Download

Additional details

Related works

Is supplemented by
Software: https://github.com/raeq/latent-semantic-recruitment (URL)

References

  • Austin, J. et al. (2021). Program Synthesis with Large Language Models. arXiv:2108.07732.
  • Barthes, R. (1970). S/Z. Éditions du Seuil.
  • Barthes, R. (1973). Le Plaisir du texte. Éditions du Seuil.
  • Chen, M. et al. (2021). Evaluating Large Language Models Trained on Code. arXiv:2107.03374.
  • Du, R. et al. (2026). Overcoming Topology Bias and Cold-Start Limitations in Drug Repurposing: A Clinical-Outcome-Aligned LLM Framework. bioRxiv. doi:10.64898/2026.01.12.699148.
  • Holtzman, A. et al. (2020). The Curious Case of Neural Text Degeneration. Proceedings of ICLR 2020.
  • Jumelet, J., Zuidema, W. & Sinclair, A. (2024). Syntactic Structural Priming in Large Language Models. Proceedings of ACL 2024.
  • Kirchenbauer, J. et al. (2023). A Watermark for Large Language Models. Proceedings of ICML 2023.
  • Krishna, K. et al. (2023). Paraphrasing Evades Detectors of AI-Generated Text, but Retrieval is an Effective Defense. Proceedings of NeurIPS 2023.
  • Kugler, R. (2025). Polysemy Patterns in Large Language Model Output. arXiv:2511.21334.
  • Leech, G. & Short, M. (2007). Style in Fiction: A Linguistic Introduction to English Fictional Prose. 2nd ed. Pearson Longman.
  • Mitchell, E. et al. (2023). DetectGPT: Zero-Shot Machine-Generated Text Detection Using Probability Curvature. Proceedings of ICML 2023.
  • Mori, M. (1970). The Uncanny Valley. Energy, 7(4), 33–35.
  • Shaib, C. et al. (2025). Detection and Measurement of AI-Generated Text Quality Dimensions: Expert Taxonomy and Span-Level Annotation. arXiv:2509.19163.
  • Stumpfe, D. & Bajorath, J. (2012). Exploring Activity Cliffs in Medicinal Chemistry. Journal of Medicinal Chemistry, 55(7), 2932–2942.
  • Tian, E. et al. (2023). GPTZero: Towards Detection of AI-Generated Text Using Zero-Shot and Supervised Methods. Preprint.
  • Vaswani, A. et al. (2017). Attention Is All You Need. Proceedings of NeurIPS 2017.
  • Wu, J. et al. (2025). A Survey on LLM-Generated Text Detection: Necessity, Methods, and Future Directions. Computational Linguistics, 51(1), 275–338.
  • Barker, P. (1991). Regeneration. Viking.
  • Berry, W. (2000). Jayber Crow. Counterpoint.
  • Bourdain, A. (2000). Kitchen Confidential. Bloomsbury.
  • Buford, B. (2006). Heat. Knopf.
  • Conrad, J. (1902). Typhoon. Heinemann.
  • Fisher, M.F.K. (1954). The Art of Eating. World Publishing.
  • Hemingway, E. (1952). The Old Man and the Sea. Scribner.
  • Hooker, R. (1968). MASH: A Novel About Three Army Doctors. Morrow.
  • Junger, S. (1997). The Perfect Storm. Norton.
  • Kesey, K. (1964). Sometimes a Great Notion. Viking.
  • McCarthy, C. (1985). Blood Meridian. Random House.
  • McPhee, J. (1975). The Survival of the Bark Canoe. Farrar, Straus and Giroux.
  • O'Brian, P. (1969). Master and Commander. Collins.
  • Orwell, G. (1933). Down and Out in Paris and London. Gollancz.
  • Pollan, M. (1997). A Place of My Own. Random House.
  • Powers, K. (2012). The Yellow Birds. Little, Brown.
  • Proulx, A. (2016). Barkskins. Scribner.
  • Remarque, E.M. (1929). All Quiet on the Western Front. Propyläen.
  • Sturt, G. (1923). The Wheelwright's Shop. Cambridge University Press.
  • Thompson, F. (1945). Lark Rise to Candleford. Oxford University Press.

Subjects

Computational Linguistics
http://id.loc.gov/authorities/subjects/sh89003285
Artificial Intelligence
http://id.loc.gov/authorities/subjects/sh85008180
Natural Language Processing
http://id.loc.gov/authorities/subjects/sh88002425