How to Assess Students’ Knowledge in the Age of AI

© iStock

A researcher at HSE University has proposed a flowchart to help lecturers decide how to assess students who use artificial intelligence. It shows where the use of AI should be restricted and where it can be incorporated into the learning process. The article has been published in IT Professional.

Artificial intelligence is transforming education, particularly in computer science. Generative AI models are used to write and debug code, solve academic problems, and explain solutions, making it more difficult to assess how independently students have mastered the material through final assignments.

A researcher at HSE University–St Petersburg has analysed current approaches to assessing students’ knowledge and, based on this analysis, proposed a decision tree (flowchart) to help lecturers select more appropriate learning activities and assessment tasks.

Flowchart for assessing students’ knowledge

 © A.A. Ternikov 

The choice of approach takes into account the type of assignment, students’ level of preparation, the possibility of automated assessment, and time constraints. For example, the author recommends placing greater restrictions on the use of AI in the early years of study.

Andrei Ternikov

‘Students need to develop fundamental knowledge in basic courses, so assessment points are best conducted with limited or no access to AI. Once students have acquired this knowledge, it is also possible to assess how thoughtfully they use AI tools,’ explained Andrei Ternikov, Associate Professor at the Department of Management at HSE University–St Petersburg and the author of the study.

At the same time, the researcher believes that a complete ban on the use of AI is inappropriate, since professionals in the field are already expected to be proficient in using AI tools. Therefore, in later years of study, the author suggests incorporating AI into assignments and separately assessing how thoughtfully students work with it. For example, students could be asked to demonstrate exactly how they used AI when completing an assignment, with the resulting work then assessed jointly by the student and lecturer.

‘This format teaches students not simply to turn to a neural network for help, but to understand its contribution, explain their decisions, and critically evaluate the result. This approach is broadly consistent with what is already being discussed and developed at the university: recommendations for working with AI and experiments being conducted at individual faculties,’ the researcher said.

At the same time, some assessment points should be completed without AI assistance, in the form of an oral examination or an assignment conducted in a controlled environment. This allows lecturers both to teach students how to work with modern tools and to check whether they have acquired the necessary knowledge and skills.

The author also noted that some of the methods had been tested on a small sample, so further research is needed to determine how well they work across different courses and universities and over the long term. The main objective remains unchanged: to ensure that assessment reflects a student’s actual knowledge and skills, even when they use AI in the learning process.