ESL ASSESSMENT IN THE AGE OF AI: SERBIAN ESL TEACHERS’ ATTITUDES TOWARDS THE USE OF CHATGPT

Following the introduction of publicly accessible AI and LLMs, they have gradually started to be adopted in the ESL classroom. Recent studies have revealed that these tools have had a wider impact on ESL assessment, allowing for enhanced accuracy, efficiency, and person alization in language evaluation. However, how widely these technologies are understood and used by teachers is still being investigated. This study explores how Serbian ESL teachers use and perceive ChatGPT as a tool for assessing student work. The study employs a mixed-methods approach, combining survey data from 53 participants with follow-up interviews from 10 educa tors to examine their tendencies, habits, and opinions regarding ChatGPT for assessment. Find ings revealed that grammar and vocabulary are the language skills which Serbian ESL teachers assessed most often using ChatGPT, while skills such as speaking, listening, and critical thinking were rarely assessed using AI. Teachers also expressed moderate trust in the tool’s ability to detect grammar and syntax errors, but were concerned about its limitations. These limitations included ChatGPT’s ability to provide accurate assessment, its tendency to give vague feedback, and its lack of contextual understanding. While many teachers recognized its ability to save time for as sessment tasks, they also recognized their own limited experiences with AI tools and stressed the need for human oversight, formal training, and ethical safeguards. This study indicated that even though ChatGPT does not yet have widespread use in the Serbian ESL classroom, interviewees frequently emphasized that it should be used as a supplementary tool to assist teachers, but that guidelines and professional training on its use need to be provided to the teachers before they would readily implement it in the classroom.