Tracing Evidential Reasoning in Scaffolded AI-Supported Collaborative EFL Argument Writing: A Classroom-Based Multi-Case Study
Journal of Language and Education · 30.06.2026 · 12 (2) · 24-36
DOI: 10.17323/jle.2026.32116 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Language and Education, online erschienen am 30.06.2026. Verfasst von Eka Fadilah. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Background
L2 argumentative writing research shows that learners often insert sources without explaining why those sources support a claim. The process through which source evaluation becomes written inferential structure remains under-specified, especially in AI-supported collaborative EFL writing.
Purpose
To trace how L2 learners evaluated curated evidence, how those evaluations became visible in Toulmin-informed argument structure, and how peer dialogue and AI-supported questioning mediated that process within a scaffolded writing sequence.
Method
The study was conducted in a 120-minute Indonesian undergraduate EFL writing session involving 22 students. A qualitatively driven multi-case design was used, with four focal dyads selected for fine-grained analysis. Product analysis was combined with interactional process analysis, structured analytic scoring, and descriptive quantification. Data included written drafts, dyadic talk, screen recordings, AI logs, and time-on-task records.
Results
Evidence-evaluation totals increased from 3–5 in pre-drafts to 8–14 in post-drafts across the four focal dyads, while Toulmin-based totals increased from 2–5 to 6–13. The largest movements appeared in warranthood contribution, argumentative function, warrant articulation, backing, and rebuttal. Across 25 AI prompts, students used AI for idea generation (n = 8; 32%), evidence checking (n = 7; 28%), reasoning/warrant testing (n = 6; 24%), and language polishing (n = 4; 16%). Inter-rater reliability was high (Cohen’s κ = .78–.86).
Conclusion
The findings offer limited analytic and pedagogical evidence. They should not be read as causal evidence of AI effects or durable development. They show how ETRE coding and the warranthood-contribution dimension can help researchers and teachers examine where evidential reasoning becomes discussable, how source judgements are carried into claims, warrants, backing, and rebuttals, and how similar scaffolded EFL tasks can be designed for critical evidence use. The study is therefore most useful for refining analytic tools and classroom design.
Zitieren
APA
Fadilah, E. (2026). Tracing Evidential Reasoning in Scaffolded AI-Supported Collaborative EFL Argument Writing: A Classroom-Based Multi-Case Study. Journal of Language and Education, 12(2), 24-36. https://doi.org/10.17323/jle.2026.32116
Vancouver
Fadilah E. Tracing Evidential Reasoning in Scaffolded AI-Supported Collaborative EFL Argument Writing: A Classroom-Based Multi-Case Study. Journal of Language and Education. 2026;12(2):24-36. doi:10.17323/jle.2026.32116
BibTeX
@article{Fadilah2026,
author = {Fadilah, Eka},
title = {Tracing Evidential Reasoning in Scaffolded AI-Supported Collaborative EFL Argument Writing: A Classroom-Based Multi-Case Study},
journal = {Journal of Language and Education},
year = {2026},
volume = {12},
number = {2},
pages = {24--36},
doi = {10.17323/jle.2026.32116}
}
EUVIMED