Several projects are currently underway at SCI to develop the use of AI in education. This page brings together examples of work related to study guidance, assessment and feedback in courses and examinations.
AI support for assessment in examinations
The project explores how AI could be used to support the assessment of examinations and submitted assignments. In a possible future process, assessment material would be retrieved from KTH’s systems, anonymised and divided by student and task. The assessment could then be carried out with the support of a language model, based on a predefined assessment rubric, with results and any feedback handled within KTH’s existing processes.
A prototype already exists and has shown promising results, but the work is still at an early stage of development. Before such a solution can be used in practice, several issues need to be investigated and addressed, for example:
how quality assurance and review of the assessment should be carried out
what type of feedback is appropriate to provide to students
how students perceive the use of AI as support in assessment
how the solution can be integrated with KTH’s systems and administrative processes
how criteria, models and instructions should be designed to produce reliable results
how a well-functioning workflow should be established, from marking to quality control
how test runs and any subsequent implementation in operational use should be carried out
what documentation, instructions and training will be required
AI-assisted grading with transparent assessment rubrics and annotated feedback integrated into Canvas. The project develops and tests AEMS (Automated Exam Marking System), a privacy-conscious, PDF-based workflow for automated grading of examinations in mathematics and engineering subjects. AEMS is fully integrated with Canvas, the learning management system used at KTH. The project has three specific objectives:
faster feedback,
transparent and reviewable assessment through clear assessment rubrics,
annotated individual feedback delivered directly within the student’s existing workflow in Canvas.
ChatGPT has been trained on course material from SK1110 Electromagnetism and Wave Motion to serve as a Swedish-language study assistant. It has helped students understand electromagnetism and waves through hints, reasoning questions and references to the course literature, rather than by providing ready-made solutions.
Going forward, similar AI agents could be developed into physics-based assistants for learning, exam design, feedback, literature reviews and interactive simulations in courses at the SCI School.
