Evidence-Grounded Decision Support for AI-Assisted Literary Interpretation: A 90-Response Exploratory Pilot Using Kashmiri Texts
Computer and Decision Making: An International Journal · 29.10.2026 · 3 (6S) · 51-63
DOI: 10.1922/comdem.v3i6s.18786 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 29.10.2026. Verfasst von Pranshu Joshi und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The application of large language models (LLMs) in the teaching of the humanities and in digital heritage and literary research presents a human decision-making challenge: when is a fluent interpretation adequate, and when does it need specialist verification? This exploratory pilot investigates a source-grounded assessment framework to support human review decisions for five selected historical English-language textual units from Kashmiri literature that have been presented in English over a historical period of time. The balanced design comprised two AI interfaces, three prompting conditions and three response-generation runs per text–interface–condition combination. Two literary questions were answered for each response. The evaluation of textual grounding, cultural specificity, literary form, interpretive ambiguity and translation awareness was carried out in one masked-identity, AI-assisted scoring pass. The mean scores for the provisional responses were 81.48/100 for Claude-assigned responses and 71.26/100 for Gemini-assigned responses, with the highest descriptive mean scores being achieved for prompts with relevant context for both interfaces. The differences were focused on literary form and ambiguity, while translation awareness was restricted. The study lacks independent expert ratings, validated acceptance thresholds, causal treatment estimates and independently verified fixed-model comparisons. Qualitative comparisons identified risks involving misattributed intention, unsupported historical generalization and unattributed religious terminology. The main contribution is a proposed human-in-the-loop decision-support workflow, linking textual evidence, claim-level uncertainty and specialist verification to responsible reuse decisions. These findings support an exploratory assessment framework; they do not establish definitive literary judgments or a validated decision-support system.
Zitieren
APA
Joshi, P., & Dave, D. N. (2026). Evidence-Grounded Decision Support for AI-Assisted Literary Interpretation: A 90-Response Exploratory Pilot Using Kashmiri Texts. Computer and Decision Making: An International Journal, 3(6S), 51-63. https://doi.org/10.1922/comdem.v3i6s.18786
Vancouver
Joshi P, Dave DN. Evidence-Grounded Decision Support for AI-Assisted Literary Interpretation: A 90-Response Exploratory Pilot Using Kashmiri Texts. Computer and Decision Making: An International Journal. 2026;3(6S):51-63. doi:10.1922/comdem.v3i6s.18786
BibTeX
@article{Joshi2026,
author = {Joshi, Pranshu and Dave, Dr. Nigam},
title = {Evidence-Grounded Decision Support for AI-Assisted Literary Interpretation: A 90-Response Exploratory Pilot Using Kashmiri Texts},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {3},
number = {6S},
pages = {51--63},
doi = {10.1922/comdem.v3i6s.18786}
}
EUVIMED