From Interaction to Improvement: English Translation and Interpretation Students’ Linguistic Growth Through Dynamic Assessment

Özturan T., Shrestha P.

7th International English for Specific Purposes Conference, Ankara, Türkiye, 6 - 07 Kasım 2026, ss.41, (Özet Bildiri)

  • Yayın Türü: Bildiri / Özet Bildiri
  • Basıldığı Şehir: Ankara
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.41
  • Erzincan Binali Yıldırım Üniversitesi Adresli: Evet

Özet

Dynamic Assessment (DA) is a learning-oriented assessment approach grounded in Vygotsky’s Sociocultural Theory (Vygotsky, 1978). Accordingly, social interaction and mediation lead to learning. For that purpose, DA blends instruction with assessment through a collaboration process between teacher and student. Instruction relies on personalized, adaptive, implicit-to-explicit feedback/instruction, and assessment focuses on present-to-future assessment that encompasses learners’ Zone of Proximal Development (ZPD) and learning potential (Lantolf & Poehner, 2004). Therefore, microgenetic and macrogenetic development are central to revealing a learner’s ZPD.

 

The rise of Generative Artificial Intelligence (GenAI) and Large Language Models has enabled teachers to implement DA in their classes because GenAI tools can understand prompts and generate relevant responses. Some studies have examined the integration of GenAI and DA in different aspects (e.g., Behbahani et al., 2026; Chen, 2026; Chuang & Wang, 2025; Wei & Shin, 2025). Yet, to our knowledge, no study has investigated how such integration might affect end-users’ actual linguistic development, a question which warrants exploration (Kushki, 2026; Yu & Poehner, 2025). Thus, this study aims to address this gap using an explanatory sequential design. 

 

Sixteen B2 CEFR-level students from an English Translation and Interpretation department at a Turkish university completed three argumentative ESP writing tasks over seven weeks. After each independent writing task, they received DA through either Gemini or ChatGPT, following a prompt piloted and validated beforehand (Authors, 2025). Written texts were analyzed for nominalisation (LancsBox X) and for text length, vocabulary sophistication, accuracy, and metadiscourse markers (Text Inspector). Statistical analysis (descriptive, Friedman, and Mann-Whitney U tests) was run in Python, and thematic analysis was conducted for interview data. 

 

Findings showed a significant increase in text length, particularly in Task 1 and Task 2. Lexical diversity improved but with no statistically significant difference. Vocabulary sophistication displayed mixed results, with C2-level vocabulary decreased before rebounded. The clearest gains were observed in metadiscourse: students used more hedges and evidential markers over time. ChatGPT-driven DA sessions drove academic vocabulary use more than Gemini-driven DA sessions. Interview data also supported these patterns.