THE 15TH SYMPOSIUM O N FOREIGN LANGUAGE WRITING · INDIVIDUAL PRESENTATION

Towards Hybrid Intelligent Feedback

in EFL Writing

A qualitative inquiry into GenAI-augmented peer dialogue and co-regulated

WENZHOU MEDICAL UNIVERSITY learning. SCHOOL O F FOREIGN LANGUAGE STUDIES

DA (ALEX) YAN, PH.D.

RESEARCH ASSOCIATE PROFESSOR · SCHOOL OF FOREIGN LANGUAGE STUDIES WENZHOU MEDICAL UNIVERSITY · YANDA@WMU.EDU.CN

Contents A qualitative inquiry into dialogic transformation and co-regulation.	1.0 2.0 3.0 4.0	Theoretical foundations and the hybrid intelligent turn Empirical methodology, platform, and qualitative corpus Qualitative findings: thematic architecture and core themes Pedagogical recommendations for writing teachers and
		Q&A

1.1	LITERATURE		REVIEW
	quality (Learner agency: Nicol, 2010).	SOCIO-CONSTRUCTIVIST Empirical potency: d = 0.73 Sustainable assessment:	Pedagogical Promise in L2 Writing	Peer Feedback: Promise and Trilemma FOUNDATION Robust meta-analytic effect size on writing, Vuogan & Li, 2023). Fosters genuine audience awareness, critical evaluation, and active sense-making (Hyland & Hyland, 2019; Decentralizes feedback from heavy teacher workloads into reciprocal peer dialogues (Wu et al., 2022).	PANADERO 2023).	& LIPNEVICH Three-Dimensional Bottlenecks Cognitive dimension: Socio-affective dimension: & Lee, 2016). (Hadwin et al., 2018).	(2022) Based on the three-dimensional framework of peer assessment Underdeveloped domain expertise leads to fixation on superficial grammar and punctuation (Latifi et al., Face-saving anxiety and peer distrust in Asian EFL contexts mute critical candor into polite praise (Yu Structural / Regulatory dimension: Feedback is consumed as static information rather than an interactive co-regulated dialogue
Da Yan · SFLS	@ Wenzhou	Medical		University			SFLS@WMU

1.2	THEORETICAL		CORNERSTONE	Interactive Versus Complementary Feedback
	combined Operates as	complementary integration feedback sources.	Overarching Theoretical Framework · Yan (2025), Computer Assisted Language Learning against interactive dialogic integration Complementary Approach Automated feedback (AWE) and peer feedback are additively or sequentially 1st-order "feedback-as-information" Students passively consume reports; zero dialogue between Fails to trigger explicit comparisons or self-feedback.	A bipartite lens discovered and proposed by the author to conceptualize how multiple feedback sources integrate in L2 writing: contrasting. Interactive Approach on drafts. the. Operates as gains.	additive Automated feedback is interactively embedded into. process of peer feedback generation. 2nd-order "feedback-on-feedback" Peers audit and refine comments via AI before sharing. Triggers Nicol's (2021) comparators & macro-organization
Da Yan ·	SFLS @ Wenzhou	Medical	University		SFLS@WMU

1.3	LITERATURE	REVIEW
Da Yan ·	SFLS @ Wenzhou	Pedagogical Framework for HIF Medical University SFLS@WMU
Da Yan ·	SFLS @ Wenzhou	Dialogic Mechanism of HIF Medical University SFLS@WMU
		HIF-Write Platform and Interaction
Da Yan ·	SFLS @ Wenzhou	Medical University SFLS@WMU

2.2	METHODOLOGY
Da Yan ·	Longitudinal Intervention Protocol SFLS @ Wenzhou Medical University SFLS@WMU

3.1	QUALITATIVE	FINDINGS
Da Yan ·	SFLS @ Wenzhou	Thematic Map of Qualitative Findings Medical University SFLS@WMU

3.2 THEME 1 · FEEDBACK PROVISION

Helping Providers Check Comments

🔍 DIAGNOSTIC VERIFICATION 💡 EVALUATIVE CONFIDENCE

Checking Accuracy Before Sharing Overcoming Insecurity & Learning Criteria

"Checking with AI prevents me from making silly mistakes before sending comments... It "In past peer reviews, I always worried: 'What if my English isn't good enough to gives me confidence that my advice is accurate." (Interview V1) critique others?' AI gave me reliable backing." (Interview P8)

"I wasn't sure if my peer's topic sentence was truly off-track. AI confirmed the logical "Asking AI to check my comments helped me see what good feedback looks like. I gap, which gave me the courage to point it out." (Interview V4) learned how to explain writing issues clearly." (Interview P3)

*"I used AI as a double-check. It showed that my comments matched the grading rubric,*Helps students write constructive, evidence-based feedback. so my peer took my feedback seriously." (Interview V2) Reduces the anxiety and self-doubt that often paralyze peer reviewers.

Da Yan· SFLS @ Wenzhou Medical University SFLS@WMU

3.3 THEME 2 · FEEDBACK RECEPTION

Helping Recipients Understand Feedback

📖 DECODING AMBIGUITY 🎯 ACTIONABLE FEED-FORWARD

Explaining What Peers Meant Giving Concrete Next Steps

"My partner wrote very brief notes. AI explained why my thesis statement was confusing, "My peer gave me a general idea, and AI showed me how to rewrite the paragraph with saving us hours of debate." (Interview E3) practical sentence structures." (Interview G1)

"When my peer wrote 'awkward transition', I didn't know how to fix it. AI explained "AI provided the missing step between spotting a problem and actually fixing it—it gave what made it awkward and gave two alternatives." (Interview E6) me real examples to work from." (Interview G3)

"AI acted like a translator. It turned cryptic peer marks into clear explanations of how "Instead of just telling me to 'use better words', AI suggested specific academic sentences connect." (Interview E1) vocabulary that fit my essay topic." (Interview G5)

3.4 THEME 3 · PEER INTERACTION

Resolving Conflicts and Reducing Anxiety

⚖️ NEUTRAL DISPUTE MEDIATION 🛡️ FACE-THREAT MITIGATION

Arbitrating Disagreements Between Peers Reducing the Fear of Critical Comments

"We noticed a problem in the essay but disagreed on how to fix it. We consulted AI "With AI helping, peer critique felt encouraging rather than embarrassing. It didn't feel together and used its advice to find common ground." (Interview M1) like an attack." (Interview S1, S2)

"When we had conflicting views, asking AI gave us an objective third opinion so we "In Chinese culture, being too direct feels impolite. AI helped me soften my comments so didn't argue or hurt our friendship." (Interview M3) my partner felt supported." (Interview S5)

"Having AI as a neutral reference prevented conversations from ending in awkward "Receiving polite, constructive suggestions made me open to revising. I didn't feel silence or one person just giving up." (Interview M5) defensive." (Interview S4)

Keeps peer discussions friendly and focused on essay quality. Fosters a safe, trusting atmosphere in the writing classroom.

3.5 THEME 4 · RECIPROCAL LEARNING

Self-Correction Through Peer Review

REFLEXIVE SELF-CORRECTION 🔄 BIDIRECTIONAL CO-REGULATION

Spotting Own Writing Flaws via Peer Review Mutual Improvement for Both Students

“After helping my peer improve her topic sentences with AI, I suddenly realized I made the “When my partner questioned what AI suggested, it made both of us rethink our writing exact same organizational mistake in my own essay! I immediately opened my draft and standards. We weren't just following AI; we were learning together.” (Interview I8) fixed it.” (Interview I2) “Feedback became a genuine two-way conversation where both of us adjusted our essays at “Critiquing my classmate's draft forced me to examine my own essay with the same critical the same time.” (Interview I11) eye. Reviewing others made me a better self-editor.” (Interview I5) Two-way benefits: The person giving feedback learns just as “Reviewing a peer's essay acted like a mirror that showed me blind spots in my own much as the person receiving it. writing that I couldn't see before.” (Interview I7) Self-reflection: Comparing peer writing against AI suggestions stimulates spontaneous self-correction on one's own draft (Nicol,

2021).

3.6		THEME	5 ·		CHALLENGES AND	CAUTIONS
	⚠️ voice." (P7)	and robotic." (V2) and edit AI wording.	ALGORITHMIC sounding like the same template." (S3)	"If you copy AI blindly, everyone's essay ends up	CLICHÉS Wordiness & Generic Phrasing "AI leaves its fingerprints... It can be wordy, repetitive, "I always rewrote AI suggestions to keep my own writing Students must remain the ultimate authors	Practical Challenges and AI Limits 🤝 EVASIVE SYCOPHANCY Excessive Politeness "When we asked AI who was right, it agreed with both of us!" (M4) "AI was too polite to take a firm stand; we had to make the final choice." (M6) "AI praised contradictory ideas. We had to check the rubric ourselves." (R4) AI tends to avoid conflict; students must exercise critical judgment.	❌ FACTUAL INACCURACIES Grammar & Hallucination Errors "AI fabricated a grammar explanation... I had to check an external dictionary." (V6) "It sounded very confident even when it was completely wrong." (E5) "We caught AI misapplying a grammar rule and corrected it together." (E2) Human checking and teacher guidance remain strictly necessary.
Da Yan ·	SFLS	@	Wenzhou	Medical	University		SFLS@WMU

4.1	TEACHING		RECOMMENDATIONS
	RECOMMENDATION Grader dialogue partner to-student discussion. own writing.	Treat AI as a Partner, Not a scoring machine. Position AI as a	1 Avoid using AI simply as an automated that supports student- Teach students to critically evaluate AI suggestions and keep ownership of their	Practical Recommendations for Teachers RECOMMENDATION 2 Teach Prompting as a Learning Strategy Integrate prompt writing into the writing syllabus. Show students how to ask AI specific diagnostic questions Prompting should help students reflect on essay structure, not just generate quick grammar fixes.	RECOMMENDATION Encourage Learning by Reviewing Peers review peer drafts with AI assistance.. Evaluating classmates' writing triggers self-reflection and self-correction students' own essays, doubling the learning benefit.	3 Design activities where students actively on
Da Yan ·	SFLS @ Wenzhou	Medical	University			SFLS@WMU

Thank you
	Questions, feedback, and inquiries are warmly welcome. Da (Alex) Yan, Ph.D. · yanda@wmu.edu.cn School of Foreign Language Studies · Wenzhou Medical University
Project Note: organization, and feedback self-efficacy have been published in Applied Linguistics (Yan, 2026).	This presentation reports on the qualitative phase of an overarching research program on hybrid intelligent writing assessment. Quantitative experimental findings regarding longitudinal writing performance gains, System (Yan et al., 2026) and International Journal of