Forum: Use AI grading tools for assignments with care
Sign up now: Get ST's newsletters delivered to your inbox
As an NUS undergraduate, I read with interest the article “Three NUS students build AI tool to tackle teachers’ marking load, give quality feedback (June 29).
I also took lecturer Lee Boon Kee’s Digital and AI Ethics course last semester.
While it may be true that artificial intelligence improves marking efficiency, this does not automatically translate to better learning outcomes. It is good that there is human review of the first cut by AI, as students do value human feedback.
This was shown by a study involving about 1,200 students at a Hong Kong university, which found that human feedback led to the highest essay score improvements, with AI feedback showing the least improvement.
The researchers found that students valued human feedback, and cited personalisation and trust as highly valued.
AI grading may diminish student-instructor trust because instructors are not transparent with how AI is used for grading.
The Institute of Electrical and Electronics Engineers strongly advocates that AI must be used with care. This is because it has the potential for algorithmic bias, and AI models often struggle to explain how they reach a specific conclusion, which is why human-in-the-loop oversight remains critical.
Educational institutions must establish clear protocols regarding AI grading. They must also be transparent about its use.
When students are left in the dark, AI grading feels opaque and we lose faith in the very mechanisms designed to measure our learning.
Terence Teo Li Yang
