Balancing AI and Teacher Professional Judgement: Swedish Upper Secondary EFL teachers’ Perspectives on Integrating Generative AI in Assessment

Sammanfattning

The integration of generative AI into educational assessment represents a significant shift in teaching practice, yet teachers' perceptions of this development remain underexplored, particularly in the Swedish context. This study investigates how Swedish upper secondary EFL teachers perceive and experience the use of generative AI in assessing students' language ability, with attention to perceived opportunities and challenges, experiences of integration, institutional support and teachers' influence on students' AI-related practices. A qualitative methodology was employed, consisting of semi-structured interviews with six upper secondary teachers in southern Sweden. The findings indicate that teachers view generative AI primarily as a support tool for repetitive and structured assessment tasks, while expressing significant caution regarding its use in summative contexts. Trust, accuracy, data privacy and academic integrity were identified as central barriers to broader AI integration, reinforcing the importance of a human-in-the-loop approach. The results further highlight that successful AI integration depends on adequate institutional support, structured professional development and the cultivation of AI literacy among both teachers and students. These findings show that AI integration in assessment is characterised by negotiation rather than replacement. While AI offers meaningful support, teachers remain essential for ensuring fairness, contextual judgement and ethical responsibility in education.