A New Era with
Artificial Intelligence Powered Learner Corpora for Teaching and Learning of
English as a Second Language in Primary School Education
Wei Lun Wong1 and Sandiway Fong2*
|
*Correspondence: sandiway@arizona.edu |
|
1Faculty of Education, Universiti Kebangsaan
Malaysia, Malaysia 2Department of Linguistics, University of Arizona, USA |
Abstract
This
innovation reports on the design, implementation and evaluation of an
AI-powered learner corpora to support English writing in a Malaysian national
primary school. The intervention targeted Year 4 to Year 6 pupils learning
English as a second language within the CEFR-aligned KSSR Semakan
curriculum, in which a significant portion struggled with vocabulary, cohesion
and sentence boundaries writing around A1 to A2 level. This work is in response
to two competing tensions: i) pupils’ increasing use
of generative AI, which produces fast but less human-involved and
locally-grounded language, and ii) the underutilisation of corpus linguistics
in primary classrooms due to technical and training demands. A 152,187 word
corpus of 560 anonymised pupil essays was compiled and uploaded into a
retrieval-augmented custom GPT. This AI-powered learner corpora provided
examples and feedback drawn only from local pupil texts and was embedded into a
three-stage writing cycle: guided drafting, AI-mediated exploration and
targeted revision. The design was informed by the Technology Acceptance Model
and the Expectation–Confirmation Model, focusing on perceived usefulness, ease
of use, satisfaction and continuance intention. Classroom observations and
teacher reflections showed that pupils became more willing to revise and
exhibited clearer control of basic sentence structure and cohesion. Teachers
reported reduced example-generation workload and more focused conferencing.
However, the impact on idea development was modest, some weaker writers engaged
in formulaic borrowing and digital literacy differences shaped access to
benefits. This paper concludes with practical recommendations for pedagogy,
equity-focused scaling and future mixed-methods research on AI-powered learner
corpora in school-based ESL writing.
Keywords: AI-powered learner corpora, English as a
second language, primary school education
1.
Teaching and Learning Context
The artificial intelligence (AI) powered learner
corpora was implemented in a Malaysian national primary school that follows the
Common European Framework of Reference (CEFR) aligned Kurikulum
Standard Sekolah Rendah
(KSSR Semakan) for English. In this curriculum,
writing is a core component. By the end of Year 6, pupils are expected to
produce short and extended texts for a range of purposes and audiences (Repo et
al., 2025). The focal classes were Year 4 to Year 6. They were approximately 10
to 12 years old. English is learned as a second language and used mainly as a
school subject rather than a home language. Most pupils spoke Malay, Mandarin
or Tamil as their first language, with English occupying a weaker position in
their language repertoires. Consistent with national evaluation studies, many
pupils in these upper primary classes were writing around CEFR A1 to A2 levels,
with frequent problems in vocabulary range, cohesion and sentence boundary
control (Nagai et al., 2020).
In routine lessons, guided and extended writing
tasks were part of continuous assessment. However, classroom observation and
teacher reports indicated that writing tasks often became highly exam-oriented
(Zeng & Huang, 2021). Teachers relied on model essays, substitution tables
and fixed sentence frames. These practices made it easier for pupils to
complete tasks but limited opportunities for genuine idea development and
experimentation. Similar tensions between curriculum expectations for process writing,
time-pressured and product-focused classroom realities have been reported in
broader discussions of CEFR implementation in the context of Malaysian primary
schools.
The school had moderate digital resources, such as
a computer lab, classroom projectors and filtered Internet access. Pupils have
some experience with online tools for projects, but generative AI and
corpus-based tools has not been systematically integrated into English writing
lessons. Teachers often described a familiar pattern of feedback fatigue, viz.,
large class sizes, frequent writing tasks and strong parental expectations made
it difficult to provide timely and individualised comments on every draft. This
mirrors wider concerns that writing teachers are expected to give more detailed
and personalised feedback than ever before, with increasing risk of burnout
(Maed, 2025).
Critically, the teaching and learning context
allowed the innovation to operationalise key constructs from the Technology
Acceptance Model (TAM). Perceived usefulness was addressed by showing pupils
that the system provided immediate and familiar examples drawn from peers’
texts. Perceived ease of use was supported through simple prompts and a
teacher-mediated interface appropriate for pupils (Jung & Jin, 2025).
At the same time, the school’s uneven access to
devices and the pupils’ varied digital literacy levels highlighted a potential
risk. Without careful design and teacher mediation, AI-powered learner corpora
could widen the divide between already advantaged pupils and weaker writers.
This tension between innovation and equity is a central concern that has shaped
the implementation of the present intervention.
2. Rationale for Innovation
The rationale for this innovation arises from two
converging trends in school-based writing. First, generative AI tools such as
ChatGPT, Grammarly and paraphrasers have become widely available and accessible
to pupils. Recent studies indicate that many EFL learners already rely on such
tools to generate ideas, reorganise sentences and correct grammar. Learners
generally report high perceived usefulness and ease of use, in line with the
TAM (Alsaedi, 2025; Alghasab, 2025).
However, there is growing concern that the
uncritical dependence on generic AI may result in to “flattened” language, weak
critical thinking and texts that do not reflect pupils’ own voices among
lower-proficiency students (Sun & Rui, 2025). Commentaries on AI-generated
essays similarly point to formulaic phrasing and the overuse of clichés. They
risk making pupils’ writing faster but less human, less situated and less
connected to local experiences (Ng et al., 2024).
For Malaysian primary pupils, this risk is serious
because they are still developing basic control of sentence structure and
vocabulary. Fully outsourcing language production to AI may undermine the
development of foundational writing skills. Second, corpus linguistics and
data-driven learning (DDL) offer powerful ways to improve writing quality by
exposing pupils to authentic patterns of language use. Corpus-based
interventions have been shown to enhance lexical diversity, lexical
sophistication and academic writing skills (Li et al., 2025). Learner corpora
have been used to identify recurrent errors, inform targeted instruction and
support curriculum design in school contexts (Bell et al., 2021). However,
these benefits remain largely unrealised in primary classrooms because corpus
tools such as AntConc and complex DDL interfaces
require specialised training and sustained teacher development (Anthony, 2016).
Corpus interfaces can be intimidating and time-consuming to use, limiting
deployment to research projects or (smaller-scale) advanced courses (Bennett
& Dhonnchadha, 2023).
The AI-powered learner corpora in this innovation
seeks to respond to both aforementioned challenges. Instead of a generic AI
model trained on adult and decontextualised texts, it integrates uploaded pupil
writing into a retrieval-augmented ChatGPT environment. This “myGPT” is then used as a classroom tool that offers
examples and feedback drawn only from the school’s own learner corpus. The
system retains the speed, interactivity and perceived usefulness that TAM
studies identify as drivers of AI adoption (Zhang et al., 2025). But it also
constrains the language resources to age-appropriate and curriculum-aligned
texts (Alsaedi, 2025; Aksakallı & Daşer, 2025). The design therefore aims to limit the
dehumanising effect of generic AI prose by foregrounding the voices and
developmental trajectories of actual pupils in the same school.
Concomittantly, the innovation lowers
technical barriers to corpus use. Traditional concordancers
are replaced with a conversational interface that can surface concordance
lines, lexical clusters and peer examples through simple prompts. This aligns
with emerging work on integrating AI with corpus methods, work that argues that
AI-driven front ends can make corpus insights more accessible to
non-specialists while preserving the methodological strengths of corpus
analysis (Anthony, 2025).
In sum, the AI-powered learner corpora represents
a concrete opportunity. It harnesses the ubiquity and speed of AI while
resisting its tendency to produce generic and less human language. By embedding
a local learner corpus inside a user-friendly AI interface, the innovation
connects corpus linguistics, TAM and sociocultural theory in a form that is
realistic for Malaysian primary ESL classrooms. It treads a middle path between
banning AI and accepting generic AI outputs. Moreover, it reframes pupils’ own texts
as the primary engine for idea generation, language development and error
repair.
3. Description of the Innovation
The innovation was designed as a full writing
cycle that connects corpus construction, AI configuration and classroom
routines. It was grounded in TAM and Expectation–Confirmation Model (ECM) with
attention to perceived usefulness, perceived ease of use, satisfaction and
continuance intention (Zhang et al., 2025). The core design principle was to
make AI fast and accessible, while ensuring that the language resources
remained local, age-appropriate and tightly aligned with the school curriculum.
The first phase focused on building the learner
corpus. Over one academic year, the researchers collected 560 essays from Year
4 to Year 6 pupils across a range of genres (narratives, descriptive texts,
simple opinion essays). After eliding names and sensitive details, texts were
cleaned for basic formatting only. Spelling and grammar errors were retained to
preserve the developmental profile of the writing. The resulting 152,187 words
corpus was then segmented into 20 text files according to year level and task
type. This structure made it easier to track which texts would be most suitable
for particular classes and assignments and it preserved the longitudinal
development of the cohort.
The second phase involved configuring the
AI-powered learner corpora in a retrieval-augmented ChatGPT 5.0 environment. 20
corpus files were uploaded into a custom GPT (a “myGPT”)
designed for internal school use only. Retrieval-augmented generation (RAG)
allowed the model to answer questions and generate examples by drawing first on
the local corpus rather than the general internet. It is similar in spirit to
how Write & Improve and Pigai use large collections of learner scripts to
drive feedback cycles (Geng et al., 2024). In practical terms, this means, upon
entering a prompt, the school corpus is searched first for relevant material,
and the system then makes use of these materials as the basis for response.
The third phase concerned classroom
implementation. Each writing unit was organised into three stages, namely,
guided drafting, AI-mediated exploration and targeted revision. During guided
drafting, the teacher introduced the topic, brainstormed ideas with pupils and
modelled simple paragraph structures on the board. Pupils then produced an
initial draft on paper or in a digital document. At this point, the teacher
introduced the AI-powered learner corpora as a “writing helper” that could show
how other pupils in the school had expressed similar ideas.
In the AI-mediated exploration stage, pupils
worked in pairs at devices. Using scaffolded prompts prepared by the teacher,
they asked the custom GPT to show “three examples of topic sentences about…” or
“short examples of how to describe a place using adjectives and prepositions.”
The system returned responses using the language of learner corpora uploaded.
When a pupil’s sentence was unclear, the teacher guided them to ask the AI for
“a better way to say my sentence using examples from our school writers.” This
process moved pupils away from copying whole AI-generated paragraphs and
towards comparing their own sentences with peer-based models, consistent with
data-driven learning principles (Anthony, 2016) and sociocultural views of
learning through mediation (Vygotsky, 1978).
The targeted revision stage translated these
examples into meaningful changes. Pupils highlighted one or two sentences in
their drafts that they wanted to improve. With teacher support, pupils used the
AI-powered learner corpora to retrieve clearer peer examples and then rewrote
their own sentences by combining their original ideas with the improved
language patterns. The teacher monitored this process closely to prevent blind
copying and to ensure that content remained the pupil’s own.
After each cycle, revised essays were again
anonymised and, where quality and ethics allowed, appended to the corpus. This
incremental enrichment created a feedback loop similar to that reported in
studies of Write & Improve and Pigai, where multiple drafting and
resubmission cycles gradually improve grammatical accuracy and complexity
(Jiang et al., 2020).
At the construct level, the immediacy of
level-appropriate examples was expected to raise perceived usefulness, while
the conversational interface and teacher-scaffolded prompts aimed to strengthen
perceived ease of use (Alghasab, 2025). The repeated
and positively reinforced use of the tool across units was designed to support
satisfaction and continuance intention in line with ECM (Bhattacherjee, 2001).
Crucially, the innovation remained teacher-led
rather than tool-led. Teachers decided which corpus segments to prioritise,
when to invite AI support, and how to frame prompts so that pupils engaged in
noticing and self-repair rather than passive acceptance. In this way, the
design sought to combine the speed and scale of AI with the pedagogical
judgement necessary to keep pupils writing authentic, developmental and locally
meaningful.
4. Reflection and Impact
Classroom observations,
pupil work and teacher reflections suggest that the AI-powered learner corpora
had a generally positive but uneven impact on pupils’ writing and on teaching
practice. Overall, pupils became more willing to revise and showed clearer
awareness of sentence boundaries and basic cohesion markers. This pattern is
consistent with recent findings that generative AI-assisted feedback can
increase revision frequency and improve writing quality when it is integrated
into structured writing cycles (Mekheimer, 2025). At
the same time, the primary school context revealed specific constraints that
temper more optimistic claims in the wider generative AI literature (Li et al.,
2025). What distinguishes the AI-powered learner corpora in this study from
generic AI writing tools is that it does not generate idealised model texts.
Instead, it uses AI to mediate access to anonymised sentences and short
paragraphs drawn from a local corpus of Malaysian upper primary ESL writing. In
contrast to conventional learner corpora, which are often static,
researcher-controlled and mainly teacher-facing, this corpus was dynamically
expanded with pupils’ revised drafts and was designed to be directly queried by
pupils during classroom writing cycles. The contribution of this study lies
precisely in showing how such a locally generated AI-powered learner corpus
operates in real, resource-constrained upper primary ESL classrooms and in
specifying the conditions under which impact is amplified or diluted.
One
clear impact was on pupils’ confidence and perceived usefulness. Many pupils
reported that seeing anonymised examples from “other children in Malaysia’s
schools” made the advice feel achievable and trustworthy. As one teacher
reflected, “When they see that these sentences come from children in other
Malaysian schools, they say, ‘I can also write like this,’ and they are less
afraid to try a second or third draft”. This echoes evidence that near-peer
models can support self-efficacy and engagement in AI-assisted writing
(Abduljawad, 2024). Compared with experiences described in studies of generic
ChatGPT use, where pupils sometimes feel disconnected from highly polished AI
text, the local corpus helped situate feedback within a familiar developmental
range. Teachers noted fewer “perfect but copied” paragraphs and more
small and targeted changes to vocabulary and sentence structure. This aligns
with learner corpus work showing that local corpora can make writing
instruction more relatable and diagnostic (Lee, 2024). The added value here is
the demonstration that a locally generated AI-powered learner corpus can
counteract the over-idealisation of AI output by foregrounding attainable peer
language. Hence, it provides a concrete design alternative to generic and
decontextualised AI feedback.
However,
the innovation did not automatically result in deeper content development.
While sentence clarity improved, several weaker writers tended to search for
short good sentences and then plugged them into their texts with minimal
adaptation. As Teacher 02 commented, “Some of my weaker pupils just hunt for
one nice sentence and paste it in. The grammar is better, but the ideas are
still very thin”. This echoes concerns in the broader AI writing literature
that tools may encourage surface-level fixes rather than substantive rethinking
of ideas (Li, 2024). In some cases, the local corpus even amplified formulaic
language already present in the cohort to reinforce safe but repetitive
patterns. This resonates with Cotos (2014) caution
that learner corpora can normalise suboptimal habits if not balanced with
richer models and explicit metalinguistic discussion. A key lesson here is that
AI-powered learner corpora must be coupled with explicit teaching about
quality, not only with retrieval of peer examples. This study therefore refines
existing theory by showing how learner corpora and generative AI can
simultaneously remediate and reproduce local norms and by identifying formulaicity drift as a specific risk in primary-level
implementation.
From
the teachers’ perspective, the tool reduced some aspects of feedback workload
and facilitated more focused conferencing. Instead of editing every essay,
teachers redirected pupils to the corpus to see several ways of expressing
similar meanings, then used class time to discuss why one option might be
clearer or more appropriate. Teacher 01 summarised this shift as follows: “I
am writing fewer long comments by hand now, but I spend more time checking how
they are using the tool and which sentences should go into the corpus”.
This hybrid approach, in which AI and teacher feedback are combined, mirrors
experimental findings that blended human–AI feedback outperforms AI feedback
alone on writing development (Mekheimer, 2025).
However, teachers also highlighted the hidden labour of curating prompts,
monitoring pupils’ interactions with the tool and deciding which revised drafts
could safely be added to the corpus. The innovation therefore prompted a
tectonic shift in the workload associated with feedback from manual correction
towards orchestration, curation and ethical gatekeeping. This reconfiguration
of labour is a substantive contribution for policymakers and school leaders who
may underestimate the ongoing human work required to sustain AI-enhanced
feedback ecosystems.
Several
unexpected outcomes emerged. First, digital literacy differences were more
salient than anticipated. Pupils who were already confident with search and
navigation quickly learned how to phrase effective prompts, while others needed
substantial handholding to interpret the AI responses. As Teacher 04 noted, “My
high-digital pupils race ahead with clever prompts, but some still struggle
just to scroll and read the responses carefully”. This reflects broader
warnings that AI integration can widen existing inequities if digital skills
and access are unevenly distributed (Li et al., 2025). Second, a small subset
of pupils initially rejected AI suggestions. They perceived them as cheating
despite the local corpus design. Teacher 05 reported that “a few pupils told
me, ‘Teacher, this is like cheating,’ so we had to talk about how using
examples is different from copying answers”. Similar patterns of
ambivalence and partial rejection have been reported in recent studies of AI
feedback in higher education (Chen et al., 2025). In the upper primary context,
this ambivalence signals the importance of early, age-appropriate induction
into ethical and productive AI use. Overall, these findings move beyond
effectiveness claims to foreground equity and acceptance as central design
concerns for AI-powered learner corpora in compulsory schooling.
Finally,
the innovation confirmed, albeit with some complication, emerging syntheses on
generative AI in L2 writing. Systematic reviews show that most gains occur when
AI is embedded in pedagogy that promotes critical engagement with feedback and
human–AI negotiation, rather than passive acceptance (Li et al., 2025; Chen,
2025). In this innovation, the strongest impact was felt when teachers actively
orchestrated discussion around retrieved examples and prompted pupils to
explain why they changed a sentence. Where AI interaction was more solitary and
unmonitored, improvements were more modest. The main lesson is that AI-powered
learner corpora can be effective in upper primary ESL writing, but only when
they are tightly integrated with teacher mediation, explicit AI literacy and
sustained attention to equity. Researchers and practitioners may therefore
refer to this study for its empirically grounded account of how a locally
generated AI-powered learner corpus reshapes revision behaviour, workload and
equity dynamics in primary ESL classrooms. For the concrete design principles,
it offers for future AI-enhanced writing interventions.
5. Future Directions
We feel future work on
AI-powered learner corpora in primary ESL classrooms should be focused in three
main directions of development (to be outlined below) with respect to pedagogy,
equity and research design. First, there is a need to develop more systematic
pedagogical frameworks that integrate AI-powered learner corpora with existing
writing curricula. Recent reviews on generative AI in language education
emphasise that impact is strongest when AI is embedded in explicit
process-writing instruction, metalinguistic talk and critical AI literacy,
rather than used as a stand-alone tool (Li et al., 2025). Future
implementations could therefore provide ready-made lesson sequences, prompt
banks and reflection worksheets that help teachers move pupils beyond sentence-level
repair to idea development, genre awareness and voice.
Second,
scaling this innovation requires sustained teacher professional development and
attention to equity. Studies on corpus-based pedagogy and AI integration show
that teachers’ beliefs, confidence, AI literary and corpus literacy strongly
shape whether tools are used critically, mechanically or not at all (Anthony,
2025; Ibrahim & Kirkpatrick, 2024). Future innovations might adopt a
collaborative design model in which teachers co-construct the learner corpus,
co-design prompts and iteratively refine classroom routines. At the same time,
any expansion must address device access, bandwidth constraints and
differential digital skills among pupils, as AI is a potential amplifier of existing
inequalities if safeguards are not built in. Low-tech variants, such as printed
concordance strips or teacher-curated corpus notebooks derived from the same
dataset could support classes with limited connectivity.
Third,
there is a need for more rigorous and diversified evaluation. Current evidence
is promising but still exploratory. Future studies should combine fine-grained
text analysis of lexical and syntactic development with measures of
self-efficacy, AI literacy and critical thinking. Moreover, we should track
effects over longer periods and across schools (Li et al., 2025). Mixed-methods
designs that include classroom ethnography and pupil interviews would help
capture how pupils negotiate authorship and ethics when writing with AI-powered
learner corpora. Extending the model to other age groups, languages, and
subjects, for example, secondary ESL, Mandarin, or content-area writing would
also test its adaptability and inform policy-level decisions on AI integration
in basic education.
References
Abduljawad, S. A. (2025). Investigating the impact
of ChatGPT as an AI tool on ESL writing. International Journal of
Computer-Assisted Language Learning and Teaching, 14(1),
1-19. https://doi.org/10.4018/ijcallt.367276
Aksakallı, C., & Daşer, Z.
(2025). Unlocking EFL learners’ insights into ChatGPT use for L2 writing: The
impacts of usage frequency and gender variations. Current Psychology, 44(9),
7957-7977. https://doi.org/10.1007/s12144-025-07437-3
Alghasab, M. (2025). Investigating EFL students’
ChatGPT usage and attitudes through the lens of the technology acceptance model
(TAM). Educational Process International Journal, 18(1),
e2025507. https://doi.org/10.22521/edupij.2025.18.507
Alsaedi, N. S. (2025). Exploring
ChatGPT’s role in EFL learning through the technology acceptance model:
Perspectives from Saudi students. Contemporary Educational Technology, 17(4),
ep594. https://doi.org/10.30935/cedtech/17302
Anthony, L. (2016). Introducing corpora and corpus tools
into the technical writing classroom through Data-Driven Learning (DDL).
In Discipline-specific writing (pp. 176-194). Routledge.
Anthony, L. (2025). Integrating AI technology
into corpus-based language learning through ChatAI. Computer
Assisted Language Learning, 1-19. https://doi.org/10.1080/09588221.2025.2589747
Bell, P., Collins, L., &
Marsden, E. (2021). Building an oral and written learner corpus of a
school programme: Methodological issues. Learner Corpus Research Meets
Second Language Acquisition, 214-242. https://doi.org/10.1017/9781108674577.011
Bennett, C., & Dhonnchadha, E. U.
(2023). Becoming corpus literate: An in-service EFL teacher education framework
for integrating corpora into EFL teaching. Applied Corpus Linguistics, 3(1),
100048. https://doi.org/10.1016/j.acorp.2023.100048
Bhattacherjee, A. (2001). Understanding
information systems continuance: An expectation-confirmation Model1. MIS
Quarterly, 25(3), 351-370. https://doi.org/10.2307/3250921
Chen, Z., Wei, W., Zhu, X., &
Yao, Y. (2025). Unpacking the rejection of L2 students toward chatgpt-generated feedback: An explanatory research. ECNU
Review of Education, 20965311241305140. https://doi.org/10.1177/20965311241305140
Cotos, E. (2014). Enhancing writing pedagogy with
learner corpus data. ReCALL, 26(2),
202-224. https://doi.org/10.1017/s0958344014000019
Geng, J., Razali, A. B.,
Ismail, L., & Noordin, N. (2024). Effectiveness of automated
writing evaluation program to enhance EFL undergraduates’ writing performance
in China. Asia Pacific Education Review, 1-16. https://doi.org/10.1007/s12564-023-09922-5
Jiang, L., Yu, S., & Wang, C.
(2020). Second language writing instructors’ feedback practice in response to
automated writing evaluation: A sociocultural perspective. System, 93, 102302. https://doi.org/10.1016/j.system.2020.102302
Jung, Y., & Jin, S.
(2025). Questioning the role of AI as collaborator: A
systematic literature review of generative AI-supported knowledge
construction. Interactive Learning Environments, 1-20. https://doi.org/10.1080/10494820.2025.2556808
Lee, J. J. (2024). Corpora, locally
sourced: An approach to addressing the specific needs of ESL writing
programs. TESOL Quarterly, 58(3), 1215-1226. https://doi.org/10.1002/tesq.3277
Li, B., Tan, Y. L., Wang, C.,
& Lowell, V. (2025). Two years of innovation: A systematic review of
empirical generative AI research in language learning and teaching. Computers
and Education: Artificial Intelligence, 9, 100445. https://doi.org/10.1016/j.caeai.2025.100445
Li, D., Noordin, N., Ismail, L.,
& Cao, D. (2025). A
systematic review of corpus-based instruction in EFL classroom. Heliyon, 11(2), e42016. https://doi.org/10.1016/j.heliyon.2025.e42016
Li, R. (2024). A “Dance of storytelling”: Dissonances between substance
and style in collaborative storytelling with AI. Computers and
Composition, 71, 102825. https://doi.org/10.1016/j.compcom.2024.102825
Mead, N. (2025). Creative righting:
Autoethnographic creative writing as a tool to prevent teacher burnout and
attrition. Teachers and Teaching, 31(7),
1100-1113. https://doi.org/10.1080/13540602.2024.2351962
Mekheimer, M. (2025). Generative AI-assisted feedback
and EFL writing: A study on proficiency, revision frequency and writing
quality. Discover Education, 4(1), 170. https://doi.org/10.1007/s44217-025-00602-7
Nagai, N., Birch, G. C., Bower, J. V., & Schmidt, M.
G. (2020). CEFR-informed
learning, teaching and assessment. Springer Singapore.
Ng, D. T., Luo, W.,
Chan, H. M., & Chu, S. K.
(2022). Using digital story writing as a pedagogy to
develop AI literacy among primary students. Computers and Education:
Artificial Intelligence, 3, 100054. https://doi.org/10.1016/j.caeai.2022.100054
Repo, E., Tarnanen, M., &
Aalto, E. (2025). Newly
arrived students’ writer agency at the end of preparatory education: Writer
profiles. International Journal of Multilingualism, 1-19. https://doi.org/10.1080/14790718.2025.2591319
Sun, J., & Rui, J. (2025).
Evaluating the longitudinal effects of AI-enhanced collaborative dialogue modes
on computational thinking and language proficiency in EFL learners: A
mixed-methods approach. Journal of Educational Computing Research,
07356331251399452. https://doi.org/10.1177/07356331251399452
Vygotsky, L. S. (1978).
Mind in society: The development of higher psychological processes.
Harvard University Press.
Zeng, J., & Huang, L. (2021).
Understanding formative assessment practice in the EFL exam-oriented context:
An application of the theory of planned behavior. Frontiers
in Psychology, 12, 774159. https://doi.org/10.3389/fpsyg.2021.774159
Zhang, N., Wong, W. L., &
Barrot, J. (2025). Primary teachers’ acceptance and sustained adoption of
AI powered learner corpora for writing instruction through TAM and ECM
perspectives. Scientific Reports, 15(1), 37399. https://doi.org/10.1038/s41598-025-
Author
bio
Dr. Wei Lun Wong is a senior lecturer at the Faculty of Education, Universiti Kebangsaan Malaysia. His research interests encompass
Teaching English as a Second Language, corpus linguistics, applied linguistics
and educational research. He has published in high-impact journals, including
Scientific Reports and SAGE Open. He actively contributes to the academic
community as a reviewer for numerous Scopus and SSCI journals. He also serves
on editorial boards and as an academic editor for reputable international
journals such as Plos One.
ORCID: https://orcid.org/0000-0002-4367-0157
E-mail: colinw@ukm.edu.my
Professor Sandiway Fong is a full professor
at Department of Linguistics, University of Arizona. His work intersects
computer science and linguistics. Principally, he is interested in the computational
modeling of linguistic theory. He is a member of the University of Arizona
Cognitive Science Programme and Second Language
Acquisition and Teaching. He has published at numerous Scopus and SSCI journals
such as Nature, Behavioral and Brain Sciences and Linguistic Variation.
ORCID: https://orcid.org/0009-0002-2874-8240
E-mail：sandiway@arizona.edu