AI and the challenges of assessing student writing : A qualitative study on how EFL teachers in a Swedish context navigate AI in student work

Sammanfattning

The study set out to investigate how a small sample of English teachers in Sweden navigate unauthorized AI use in students’ written work and how they try to ensure that students get fairly assessed on their actual language ability. As artificial intelligence (AI) is becoming increasingly prevalent in education and continues to develop rapidly, further research is needed on how teachers navigate AI-related challenges in assessment and how they ensure that students are assessed fairly based on their actual language ability. The study is qualitative and had semi-structured interviews as its main method together with content analysis of the transcripts. All of the six participants taught English at the secondary level, whereof three were lower secondary school teachers, and three were upper secondary school teachers. The teachers were asked about their general work experiences with AI, how they ensure that a student’s written work reflects their language ability, what strategies they have if they identify AI use, and how they can counteract unauthorized AI use in written work. Results showed that teachers usually focus on having a holistic view when assessing students, arguing that most of the time students’ communicative ability goes hand in hand throughout different mediums such as writing, talking and discussing. The most common tool to counteract unauthorized AI use was through exam browsers such as Inspera, which locks down the students’ Ipads or computers during examinations. However, some teachers reported that there have been instances of students who were able to bypass the system through different forms of hacking or technical loopholes. These findings suggest that although exam tools, such as Inspera, are widely used to reduce unauthorized AI use, teachers still perceive challenges related to ensuring valid assessment and preventing students from exploiting technical loopholes.