Truace
Subscribe

From ChatGPT to Classroom Practice: A Visualized Protocol for Generative AI-Supported Feedback, Critical Thinking, and EFL Writing Instruction

This study examined associations between Visualized Protocol-guided ChatGPT Writing Instruction (VP-CGWIP), writing-score trajectories, critical-thinking dispositions, and post-intervention feedback judgment in an English as a foreign language (EFL) course. Two intact second-year university classes completed a ten-week intervention and four-week follow-up. The complete-case sample included 43 experimental and 42 control participants. Writing was assessed using the International English Language Testing System (I…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1448· cite & verify
From ChatGPT to Classroom Practice: A Visualized Protocol for Generative AI-Supported Feedback, Critical Thinking, and EFL Writing Instruction
"Dzaleu Students Learning Ewondo Language Classroom" by Clemms78 is licensed under CC BY-SA 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-sa/4.0/.

In brief

Researchers compared two intact second-year university classes in an English as a foreign language course. One class received writing instruction guided by a visualized protocol that incorporated ChatGPT feedback, and the other served as control. The intervention ran ten weeks, followed by a four-week follow-up, with 43 experimental and 42 control students in the complete-case sample.

The results suggest that structured use of a generative AI tool in writing classes may be associated with lasting writing improvement and better judgment of feedback. The design limits how far that can be pushed: each condition was a single class, the findings are labeled exploratory, and most critical-thinking disposition dimensions showed no adjusted effect, so causal claims remain unsettled.

Main points

  1. Writing was scored using IELTS Task 2 criteria, and the experimental group's overall-writing change exceeded the control group's by 0.48 points at the immediate posttest and 0.44 at the delayed posttest.
  2. The overall score and all four writing dimensions had interaction p values below 0.05 after Holm adjustment across 18 tests.
  3. Among seven critical-thinking disposition dimensions, only analyticity passed the adjusted threshold, along with the total score.
  4. Feedback judgment accuracy was higher in the experimental class at both posttests, but perception-gain correlations did not pass adjusted significance.

The gain

University EFL students taught with a structured ChatGPT protocol showed larger IELTS-style writing score gains than a control class, and the advantage was still present four weeks after the intervention.

The rundown

The study tracked writing-score trajectories across an immediate posttest and a delayed posttest four weeks later. The reported confidence intervals for the experimental-minus-control difference were 0.26-0.70 at T1 and 0.22-0.66 at T2, showing the gap stayed similar in size.

Beyond writing scores, the team measured critical-thinking dispositions using the Chinese version of the Critical Thinking Disposition Inventory. Only analyticity, plus the inventory's total score, showed an adjusted interaction effect, so the evidence for broad disposition change is limited.

A feedback judgment task was given only at the two posttests, so no baseline exists for it. Logs and journal summaries were used for descriptive context rather than as tested outcomes.

Sources

  1. Peer-reviewedJournal of Visualized Experiments2026-10-09

The debate