Exploring the impact of technology on English as a second language writing: linking outcomes, student engagement, and teacher guidance

Nguyen, Hoai Ly Nga (2026). Exploring the impact of technology on English as a second language writing: linking outcomes, student engagement, and teacher guidance. University of Birmingham. Ph.D.

Nguyen2026PhD_Redacted.pdf
Text - Redacted Version
Available under License All rights reserved.
Download (8MB)

Abstract

This study investigates the impact of four digital writing tools—Sketch Engine (corpus), Google Translate (machine translation), QuillBot (automated paraphrasing), and Grammarly (automated writing evaluation)—on English as a second language (ESL) university students’ IELTS Academic Writing Task 2 performance, tool-use strategies, and perceptions. It also examines the extent to which students’ actual tool-use practices align with teachers’ recommended strategies and explores how behavioural, cognitive, and emotional engagement (“critical engagement”) mediates writing outcomes. Twenty ESL Vietnamese university students, with various levels of English proficiency, completed two timed IELTS Task 2 essays: Essay 1 without tools and Essay 2 with access to all four tools, followed by semi-structured interviews. Tool use strategies of all 20 students were directly observed and screen-recorded by the researcher. Essays were rated by three independent raters using the IELTS rubric. Ten ESL teachers provided lesson plans with recommendations of tool usage and participated in one-on-one interviews with the researcher.

Quantitative findings showed a statistically significant improvement in overall writing performance following tool use: 15 of 20 students increased their total scores, with the mean rising from 6.48 to 6.87 (p = .0026). Significant changes were observed in Lexical Resource (LR), and Grammatical Range and Accuracy (GA) criteria, but not in Coherence and Cohesion (CC) and Task Response (TR). Among the four tools, only Sketch Engine exhibited a moderate positive association with Essay 2 scores (r = .529, p = .016).

Qualitative findings further suggested that perceived language learning gains were shaped less by tool access or frequency of use than by students’ cognitive engagement among higher-scoring students (mean Essay 2 score ≥ 6.5). Higher scorers showed stronger cognitive engagement by using multiple tools strategically, cross-checking and rejecting inaccurate outputs, and demonstrating awareness of tool limitations. They also reflected more deeply on vocabulary, collocation, and error noticing as learning opportunities during interviews. In contrast, lower scorers (mean Essay 2 score < 6.5) tended to rely on a single tool for basic functions and accepted suggestions with minimal contextual evaluation, showing weaker evidence of noticing and revision. Case study analyses further indicated that score declines were often associated with uncritical adoption of tool outputs, producing awkward phrasing or errors, including occasional small declines among higher-proficiency learners when feedback missed subtle lexical or coherence problems.

Notable gaps were identified between teacher guidance and student practice. Teachers generally promoted critical evaluation and assigned stage-specific roles to tools (e.g., Google Translate and Sketch Engine for planning; QuillBot for composing; Grammarly for revising). However, many students bypassed planning and concentrated tool use in composing and revision, with Google Translate and Grammarly dominating overall usage. Alignment with teacher-recommended multi-tool strategies was most consistently observed among high scorers, who also extended strategies (e.g., back-translation) in flexible ways, whereas low scorers more often used tools as shortcuts, constraining language development even when scores improved.

The thesis contributes theoretically by linking engagement with Output and Noticing theories, showing that language learning depends on how learners actively produce texts, evaluate and refine tool outputs, rather than access to technology. Empirically, it demonstrates that technology access alone does not ensure learning, identifies an ethics gap regarding plagiarism and overreliance, and notes the growing convergence of tool functions as Generative AI evolves. Pedagogically, it calls for explicit training in cognitive engagement for tool use, proficiency-sensitive scaffolding, and ethical awareness of Generative AI use for essay writing. Methodologically, it shows the value of triangulating multiple data sources to capture behavioural, cognitive, and emotional engagement. The study concludes that digital tools can enhance L2 writing and L2 learning only when learners engage with them strategically, positioning technology as support rather than substitute.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):