人工智慧寫作助理對於英文學習者的寫作之成效研究

Abstract

對於許多EFL高中學生來說，英語寫作可能是一個重大挑戰，他們因而尋求線上寫作工具的支援。如今，隨著科技的進步，有各種AI寫作助理可供語言學習者使用，以檢查和評估他們的寫作。鑑於這些AI寫作助理在英語寫作中的重要作用，本研究比較了兩種以ChatGPT驅動的寫作助理對EFL高中生的英語寫作的有效性，分別是Cool English Writing Assistant和Cool English AI Grammar Checker。此外，本研究旨在調查高中學生對這兩種寫作助理使用的看法。本研究實施了前測和後測以收集量化數據，並發放問卷以探究質性資料。研究對象為來自臺灣南部一所高中的三個班級的112名學生，她們被招募來使用這兩種寫作助理來評估她們的文章。學生被分為兩組(73人和39人)，並分別使用Cool English Writing Assistant和Cool English AI Grammar Checker來檢測她們的寫作。在前測中，受試者被要求寫兩篇文章，這些文章由兩位高中老師評分。在實驗前，受試者接受這兩個寫作助理的使用教程。在實驗期間，參與者寫了五篇文章並使用這兩個工具檢查和修改他們的作品。在後測中，受試者以與前測相同的兩個題目寫了另外兩篇文章，這些文章也由前測相同的兩位老師評分。之後，參與者填寫了問卷以調查她們對於這兩個寫作助理的看法。經過分析受試者的成績後，統計結果表明Cool English Writing Assistant和Cool English AI Grammar Checker皆有助於高中學生的英語寫作技能。使用Cool English Writing Assistant的學生在組內的後測成績上有顯著地提高。然而，兩組受試者在後測成績上沒有顯著差異，這表明這兩個工具對高中學生的影響相似且同樣有用。根據問卷的結果，兩組受試者都認為這些工具在檢查他們的英語寫作方面是有用且有效的。她們認為這些工具的使用者介面是友善的，也稱讚其易用性以及對詞彙使用、句子和搭配詞修改上的幫助。這些都提高了她們對英語文法和寫作技巧的理解。許多學生表示願意向朋友推薦這些工具。然而，有學生指出處理速度和偶爾的系統故障的問題。對於兩個寫作助理的優點，受試者提出如友善的介面、在特定領域的有效性、操作的簡便性、有幫助的寫作建議和快速的處理速度。缺點則包括較慢的處理速度、登入過程較繁瑣、字數限制、改變受試者的句子原意及在使用前需要將手寫文章內容數位化的過程。最後，本研究根據結果提出了研究的教學意義和侷限性，並對未來的研究提出了一些建議。For many EFL high school students, English writing can be a major challenge, prompting them to seek online writing support. The growing trend on writing assistants and technological advancements has led to the development and widespread use of various grammar-checking tools. Nowadays, with the advance of technology, various kinds of AI-powered writing assistants are available for language learners to check and evaluate their writing. Given the significant role of these AI-powered writing assistants in terms of English writing, this study compared the effectiveness of two ChatGPT-based writing assistants on EFL high school students’ English writing, namely Cool English Writing Assistant and Cool English AI Grammar Checker. In addition, this study aimed to investigate the perception of high school students toward the use of these two writing assistants.In the present study, a pretest and posttest were conducted to collect the quantitative data, and questionnaires were then distributed to probe into the qualitative information. A total of 112 participants from three intact classes at a high school in southern Taiwan were recruited to apply these two writing assistants to the evaluation of their writing. The students were divided into two groups (n=73 and 39) and were assigned to use Cool English Writing Assistant and Cool English AI Grammar Checker to evaluate their writing respectively. In the pretest, participants were asked to write two essays, which were then graded by two high school teachers. Participants were then given a tutorial on how to utilize the two target writing assistants to check their writing before the experiment. During the experiment, the participants wrote five essays and used the two tools to check and revise their works. In the posttest, participants wrote two essays on the same two topics as the pretest, and the essays were also graded by the same two teachers in the pretest. Afterwards, a questionnaire was filled out by the participants to research into their perceptions toward the application of the two writing assistants.After the grades were analyzed, the statistical results indicated that Cool English Writing Assistant and Cool English AI Grammar Checker can benefit high school students' English writing skills. Students using the Cool English Writing Assistant showed significant improvement in their posttest scores compared to their pretest scores. However, there was no significant difference in posttest performance between the two groups, suggesting that both tools have a similar impact and are equally useful for high school students. According to the questionnaire results, students from both groups found the tools useful and effective for checking their English writing. Theypraised the tools for their user-friendly interfaces, ease of use, and helpful word, sentence, and collocation revisions, which improved their understanding of English grammar and writing techniques. Many students expressed a willingness to recommend these tools to friends. However, issues such as processing speed and occasional system failures were noted. Participants also highlighted strengths such as user-friendly interfaces, effectiveness in specific areas, ease of operation, helpful writing suggestions, and fast processing speeds. Weaknesses identified included low processing speed, cumbersome login processes, word count limits, unintended changes in sentence meanings, and the need to digitize handwriting before using the tools. Lastly, pedagogical implications and limitations of the study were yielded based on the analysis of the findings. Additionally, suggestions for future research were presented.