本研究從傳統的問題解決或錯誤修正，轉向培養學習者在英語作為外語（EFL）寫作中，針對關鍵問題進行發現、探究、探索與解決的能力。研究旨在使學習者意識到自身在寫作中的主要與次要問題，進而集中精力改進。同時，本研究實施了 SQACR 策略，一個包含總結（Summarize）、提問（Question）、回答（Answer）、釐清（Clarification）與修訂（Revision）的五步驟引導流程，並在結合情境化與個人化（CP）的環境中，運用生成式人工智慧（GAI）協助學習者理解並修正其問題。我們開發了名為「Smart Roamlingo」的行動應用程式，透過影像轉文字辨識（ITR）技術促進情境化，並結合客製化微調模型，以基於個人化的 SQACR 策略提升學習者對錯誤模式的覺察與反思。在為期 9 週的準實驗研究中，將 86 名大學生分為實驗組（EG）與控制組（CG），並根據其外部 AI 使用情況分為「AI 正確使用組（PAI）」與「AI 過度依賴組（OAI）」。結果顯示，EG-PAI 組顯著優於其他組別，而 OAI 使用則對兩組學生皆具負面預測力，證實過度依賴會減損教學效益。此外，EG-PAI 組在組織構面等高階錯誤的修正上更為有效，顯示找出主次要錯誤模式有助於 EFL 學習者反思並察覺寫作中的關鍵問題。總結，SQACR 策略成功推動學習者從被動解決問題，轉變為主動發現、探究、探索與解決。;Shifting from conventional problem solving or error corrections to fostering learners′ skills in critical problem finding， enquiring， exploring， and solving， this study tried to foster English as a Foreign Language (EFL) learners to have the awareness of their major/minor problems in EFL writing. Afterwards， they can focus their efforts to correct problems and get good improvement. Meanwhile， the SQACR strategy， a guided five-step process of Summarize， Question， Answer， Clarification， and Revision， with Contextualization and Personalization (CP) environment supported and Generative Artificial Intelligence (GAI) support was also implemented to guide learners how to understand and correct their major/minor problems. We have developed a mobile app namely Smart Roamlingo， that facilitates contextualization through Image-to-Text Recognition (ITR)， powered by a custom fine-tuned model to foster EFL learners′ awareness and reflection on their error/problem patterns based on the SQACR strategy. In a 9-week quasi-experiment， 86 university students in Experimental (EG) and Control (CG) groups were categorized as Proper Use of AI (PAI) or Overreliance on AI (OAI) based on external AI usage. The results showed the EG-PAI subgroup significantly outperformed all others， while OAI usage was a negative predictor for both EG and CG students， confirming overreliance undermines pedagogical benefits. Furthermore， the EG-PAI group was also more effective at reducing higher order errors like organization dimension， which demonstrating major/minor error patterns finding was indeed useful for EFL learners to reflect and aware their critical errors or problems in EFL writing. In addition， the SQACR strategy， which can reflective process successfully moves EFL learners from passive problem solving to active problem finding， enquiring， exploring， and solving， highlighting a design direction for AI tools that fosters deeper learning.
