Measuring Writing Identity in AI-Mediated ESP Contexts: A Case Study from China

Abstract

Aims: To propose and pilot-test a mixed-methods protocol for measuring three dimensions of writing identity—authorial ownership, AI dependency perception, and perceived legitimacy—among adult ESP learners using generative AI tools. Study Design: An exploratory sequential mixed-methods design integrating quantitative scales, usage logs, text-trace analysis, and qualitative interviews through triangulation. Place and Duration of Study: A four-week English course at a university in China. Methodology: Twelve adult ESP learners (7 female and 5 males; aged 22–25 years) completed three writing tasks over four weeks. Participants completed the WIS-AI scale before and after the intervention, maintained AI usage logs, submitted successive drafts for trace analysis, and participated in semi-structured interviews. Quantitative data were analysed descriptively, and qualitative data were analysed thematically. Results: The protocol was feasible, with 10 of the 12 participants completing all measures. Authorial ownership declined slightly (3.4 to 3.1), AI dependency perception remained stable (3.8 to 3.7), and perceived legitimacy increased marginally (3.2 to 3.5). The mean suggestion acceptance rate was 67%, and grammar corrections were accepted most frequently (81%). Interview themes included an “AI mirror” effect, authorship-negotiation strategies, and shifting legitimacy standards. The pilot findings informed revisions to the scale items, interview probes, and logging procedures. Conclusion: The protocol provides a theoretically grounded approach to measuring writing identity in AI-mediated ESP contexts. The preliminary findings indicate complex relationships between AI use and learners’ identity perceptions. Although validation with larger and more diverse samples is required, the protocol offers a practical template for researchers and a diagnostic tool for practitioners.