CoChat

AI-Integrated Scaffolding to Enhance Agency and Creativity in K-12 English Language Learners: A Systematic Review

Open access
Li, Molly, Wilson, Joshua · 2025 · Information

The DOI matches its Crossref record; 43 of the 43 references it cites were found in a registry or an index; 54 of the 85 sources its claims cite were checked for support.

Abstract

This systematic literature review examines how artificial intelligence (AI)-integrated tools are being used to scaffold writing instruction for K-12 English as a Foreign Language (EFL) students, with a particular focus on preserving student agency and creativity. Drawing from sociocultural and cognitive theories of writing, this study synthesizes findings from 14 peer-reviewed empirical studies published between 2018 and 2024. Studies were analyzed using a thematic synthesis approach to identify how AI tools function as cognitive supports, creative collaborators, and language enhancement systems. The results reveal that AI tools, when implemented thoughtfully, can reduce cognitive load, foster ideation, and support self-regulated learning without undermining student autonomy. Three major categories of AI scaffolding emerged, i.e., cognitive support tools (e.g., grammar correction, idea organization), creative support systems (e.g., story generation, narrative feedback), and language enhancement tools (e.g., vocabulary expansion, stylistic refinement). Key themes include the importance of balancing AI support with student decision making, aligning tool use with theoretical models, and ensuring equitable access. Limitations in the current literature include short-term study durations, a lack of standardized metrics for creativity and agency, and the predominance of research in EFL contexts. Future research should expand on ESL settings, adopt rigorous methodologies, and explore long-term impacts on writing development.

Overview

CoChat Summary

TL;DR

This systematic review synthesizes 14 peer-reviewed empirical studies (2018–2024) on AI-integrated scaffolding in K-12 EFL writing, finding that AI can reduce cognitive load, foster ideation, and support self-regulated learning while preserving agency when implemented thoughtfully. Effects on agency and creativity depend on instructional design, tool affordances, and learner interaction, with dialogic/customizable tools generally outperforming static correction-focused tools.

Key findings

  • Three AI scaffolding categories emerged: cognitive support tools (e.g., grammar correction, idea organization), creative support systems (e.g., story generation, narrative feedback), and language enhancement tools (e.g., vocabulary expansion, stylistic refinement).
  • AI tools reduced cognitive load and supported ideation/self-regulated learning without necessarily undermining autonomy; e.g., Song and Song (50 12th-grade students in China) found improved writing skills and motivation, while Woo et al. (23 secondary students in Hong Kong) found students made strategic decisions about AI-generated text.
  • Dialog-based AI systems and tools with customizable prompting interfaces were more effective in fostering creative expression and autonomy than static, correction-focused tools; some studies reported risks of over-reliance and limited transformative impact when scaffolds were overly prescriptive.
  • PRISMA selection: 91 records identified; after removing 21 duplicates and 4 unrelated records, exclusions included 6 outside the date range, 23 not K-12, and 29 lacking relevance to AI writing instruction, yielding 14 studies; 2023 had eight studies and 2024 had four as of November.

Methods

  • Systematic literature review following PRISMA guidelines; inclusion criteria: K-12 students/teachers in EFL contexts, AI tools for writing instruction, outcomes related to cognitive engagement/creativity/scaffolding/autonomy/teacher perspectives, published 2018–2024, English.
  • Searched Education Source, ERIC, and Web of Science using keywords such as “AI in writing education,” “AI scaffolding in writing,” “Natural Language Generation (NLG) in education,” “EFL writing,” “student agency with AI tools,” and “cognitive engagement in AI writing tools”; final set was 14 empirical peer-reviewed studies.
  • Thematic synthesis conducted manually (no software); first author independently coded all studies and initial synthesis, themes discussed with second author; studies clustered by AI tool types and scaffolding strategies into cognitive support, creative support, and language enhancement categories; no risk-of-bias assessment or effect-size synthesis.

Limitations

  • Stated: many studies used short-term implementations (typically one semester or less), relied heavily on self-reported data, and few employed rigorous experimental/quasi-experimental designs, so causal links cannot be established; key constructs like agency and creativity were inconsistently operationalized with no standardized metrics.
  • Stated for the review itself: inclusion limited to K-12 EFL contexts and English-language publications, limiting generalizability to ESL/L1 settings and possibly missing relevant non-English research; no formal study-quality/risk-of-bias appraisal; excluded AI-supported reading/speaking studies; rapid AI change may date findings.
  • Evident from the text: some included study details appear inconsistent or misaligned (e.g., Coenen et al. is described in the text as examining professional writers, while Table 1 lists 8 middle school students in China), and some tools/studies may not clearly fit the stated K-12 EFL AI-writing inclusion criteria (e.g., digital mind mapping in Karim and Mustapha).

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/info16070519

Cites

43 of 43 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

45 Valid · 8 Not valid · 27 Not checked · 87 Own results and interpretations

79 more claims · 19 with their cited sources checked

Create a free account to see all 167 claims and to verify the 11 references still open.

See all claims — free account
  • Probably not a substantial claim. 46
  • It may state two separate claims as one. 48
  • The quote only partly states this claim. 4

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

3 of 3

Citations

Cites

Research works printed but not linked: 2

  1. Chu, 2022, Roles and research trends of artificial intelligence in higher education: A systematic review of the top 50 most-cited articles, Australas. J. Educ. Technol., 38, 22
  2. Cole, M., Jolm-Steiner, V., Scribner, S., and Souberman, E. (1978). Mind in Society: Development of Higher Psychological Processes, Harvard University Press.

Cited by

See all 49

Metadata

Authors
Li, M., & Wilson, J.
Affiliations
University of Delaware (US)
Corresponding author
Li, Molly
Venue
Information
Publisher
MDPI AG
Published
2025-06-21
Volume
16
Issue
7
Pages
519
ISSN
2078-2489 (ISSN-L)
OpenAlex
W4411565464
Cited by
49
References
38
Citations by year
2026: 40 · 2025: 9
FWCI
34.71
Type
journal article
Language
en
Open access
Locations
  • Information (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Social Sciences › Arts and Humanities › Language and Linguistics › EFL/ESL Teaching and Learning
  • Social Sciences › Education › Artificial Intelligence in Education
  • Social Sciences › Education › English Language Learning and Teaching
Keywords
AI scaffolding, writing instruction, English as a foreign language, English language learners, student agency, creativity, K-12 education, cognitive load, student self-regulation, grammatical error correction, story generation, vocabulary enrichment

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.