Comparison of the qualities of written texts of Turkish as a foreign language learners and texts written by ChatGPT

Today, artificial intelligence-powered language models are increasingly being used in language teaching processes. ChatGPT works by taking user inputs such as written or spoken sentences and words and converting them into structured language commands. These commands are then sent through an artificial intelligence learning algorithm that can recognize the grammar and structure of the language, as well as understand and respond appropriately in context. Research highlights many positive features of using ChatGPT in foreign language teaching and in developing speaking and writing skills. However, it is also noted that users should use it with an ethical understanding and approach the information obtained critically. It is necessary to investigate how artificial intelligence can be used in language teaching environments and whether it provides reliable information. The purpose of this study is to determine whether ChatGPT can be used to improve writing skills in teaching Turkish to foreigners and to compare the written expression skills of individuals learning Turkish as a foreign language with the texts generated by ChatGPT in terms of complexity, accuracy, fluency, and functional adequacy. The study was conducted using the document review method, one of the qualitative research methods. Within the scope of the research, the texts written by 43 students learning Turkish at the A2 and B2 levels on three different topics according to the instructions in the textbook were examined. Then, one text on the same topics was created by the researcher using ChatGPT with the same instructions. The resulting texts were analyzed using complexity, accuracy, fluency criteria, and a functional proficiency rating scale. According to the findings, ChatGPT texts generally have higher rates of complex sentences in terms of complexity, but this rate is not consistent across levels. In terms of accuracy, ChatGPT has significantly higher accuracy rates than student texts at both the sentence and word levels. In terms of fluency, ChatGPT's averages were found to be higher in terms of different word usage, but student texts achieved higher values in some areas in terms of meaningful word ratio. In terms of functional adequacy, ChatGPT texts scored higher than student texts, particularly in terms of task requirements and consistency criteria. However, it was also noted that some A2-level student texts were of higher quality than B2-level texts. These results show that ChatGPT can provide support in improving sentence and word accuracy in terms of grammar and ensuring the functional adequacy of the text in the process of developing writing skills in teaching Turkish as a foreign language. Functional adequacy encompasses presenting content that is sufficient and relevant to the topic in terms of meaning, meeting task requirements, writing texts with clearly understood purposes and ideas, and making the text a coherent whole with transition and connection expressions. It is recommended to use ChatGPT during the text planning or review stages to support learners of Turkish as a foreign language in organizing content that is grammatically correct and contextually appropriate.