Using Al to Grade Assignments#
Important note: AI grading should be approached cautiously and is best suited for formative assessment and interim feedback rather than summative grading.
AI can speed up grading, apply rubrics more consistently, and give richer feedback than most teachers can manage in one sitting. However, AI grading is still rare in classrooms. Current tools require oversight, can miss nuance, and raise privacy concerns when student work is uploaded to third-party platforms. Hence, for most teachers, using AI to grade everything would still be more work than benefit.
In the future, universities or academic institutions could run secure local AI systems to grade work without sending data outside the institution. AI could handle the first pass: applying detailed rubrics, checking requirements, and producing draft feedback. Teachers would then validate and adjust, focusing attention where human judgment matters most.
At the same time, in the future assignments will shift toward process-focused assessment. Therefore with personal AI assistants constant grading becomes less important. Instead, AI can track student progress and provide ongoing feedback ensuring every students has personalised attention until the content is mastered. Hence these AI assistants can provide personal help while the teachers can focus on the meaningful human connections, experiences and big picture learning.
How can you already begin using AI for grading today#
The following show varying degrees as to how teachers can already begin using AI today to help with grading.
A. Grammar, clarity and plagiarism checks – Let AI flag errors so teacher time is spent on higher-level thinking and content quality.
B. Rubric-based feedback drafting – Give your own scores or notes, and have AI turn them into full, clear feedback paragraphs for students.
C. Marking and pattern spotting – AI can assess answers, explain correct/incorrect responses, and identify common mistakes to focus on in lessons.
D. Feedback checkpoints for students – Have checkpoints in assignments where students upload work to an AI and get custom feedback. They can improve their work before submitting it for teacher review, and discuss changes made.
Practical example#
Professor Bas Haring (Leiden University) experimented with having student Alicia Cai use only AI tools (such as ChatGPT and Claude) for thesis supervision. AI excelled at technical guidance (eg. providing code, detecting errors, and offering conceptual explanations) and was available 24/7 to improve her academic writing. However, the student reported feeling isolated without human accountability, missed the theoretical frameworks and critical push from an experienced supervisor, and lacked personal introduction to the research community. The thesis received a grade of 8.5, with both supervisors assessing it using standard criteria. This demonstrates that AI-supported content feedback can work well in practice with current systems, though emotional and social support, along with deeper theoretical guidance, remain essential human roles. Hence AI works best for interim feedback throughout the process rather than replacing comprehensive human supervision.
