2026_19_4_13
Open access Issues > Issue n_4_2026 > Issue n_4_2026
From Teacher to AI-Generated Multimodal Feedback – Investigating the Role of Feedback Agent in Academic Writing Instruction
Izabela Olszak – Jarosław Krajka
DOI: 10.18355/XL.2026.19.04.13
Abstract
This article compares the effectiveness of AI-generated written corrective feedback and teacher-provided multimodal feedback, combining written, voice, and facial cues in an academic writing context. While the integration of AI in educational settings continues to garner increasing attention, empirical investigations that contrast the depth and impact of human versus AI-mediated feedback remain scarce. Most existing studies focus on higher-level learners, leaving the needs of lower-proficiency academic English writers underexplored. This study seeks to evaluate not only the accuracy of each feedback modality but also its influence on learner outcomes, particularly in terms of writing development and perceived pedagogical value. The research involved 40 undergraduate students, divided into two groups: one receiving teacher-provided multimodal electronic feedback (written, voice, and facial) during structured workshops, and the other receiving AI-assisted feedback generated by ChatGPT after prior training by an instructor. The study first scrutinized current Large Language Models for capacity to deliver multimodal feedback, thereby addressing questions of how multimodal AI feedback can be and how it parallels or diverges from human multimodality. By assessing revision quality, learner perception and writing progress over time, this study advocates for a more conscious integration of teacher-generated and AI-generated feedback into a carefully designed writing pedagogy.
Key words: WCF, AI in L2 learning, Writing pedagogy, Multimodal feedback
Pages: 228-240
Full Text