AUGUST 3 — In many Malaysian universities, getting an A- or even a B+ seems to have become the new abnormality, often followed by student emails asking why they fell short of a perfect A. Whether this reflects smarter students, grade inflation or something else entirely is open to debate. But one thing has become increasingly difficult to ignore come marking season.
Many of my colleagues and I now find ourselves spending hours grading essays and theses that are strongly suspected of having been largely generated by artificial intelligence (AI). It raises a troubling question: who, exactly, are we assessing?
AI is indeed one of the most transformative educational tools of our time. But in embracing these technologies, we must pause and also ask whether we have unintentionally allowed AI to become more than just a tool.
For decades, educators have relied on Bloom's Taxonomy to understand learning. For instance, in an English language classroom, students first remember vocabulary and grammar rules and understand how they function. They then apply this knowledge to new contexts, analyse texts, evaluate language choices and finally reach the peak of the mountain by creating an original essay. Traditionally, that final piece of writing has been seen as evidence of deep learning.
Today, however, "creation" can be reached with a simple prompt entered into a large language model (LLM). Within seconds, a polished essay appears.
The summit has become the baseline.
When a large portion of the cognitive work can be done by an algorithm, an essay no longer necessarily equates to deep learning. This has left educators caught in a gruelling game of cat and mouse. Many rely on AI detectors whose reliability remains questionable, while students wonder why they are discouraged from using AI to write when lecturers are rumoured to use it to grade.
Whether or not that comparison is entirely fair, it exposes a deeper issue: we have become engrossed in policing AI instead of redesigning assessment. If we continue assessing only the finished product, we risk grading the AI as much as the student.
An idea gaining traction in AI education is the concept of a reversed Bloom’s Taxonomy, popularised by educator Michelle Kassorla. The idea is that if AI can now produce what was once considered the highest demonstration of learning, then perhaps the real proof of learning begins only after the AI generates the first draft.
This means, for example, asking students whether they can identify factual inaccuracies or hallucinations. Can they critique weak reasoning or challenge unsupported claims? Furthermore, can they verbally defend those revisions and explain the thinking behind every decision they made?
In this inverted mountain, students are no longer assessed on their ability to produce text, but on their ability to truly know, unpack and own the information. Deep understanding and internalised knowledge now stand at the peak of human learning.
For Malaysian lecturers, this should prompt us to rethink assessment rather than fighting against technology. We should require students to demonstrate what AI cannot do. This means shifting attention from polished products to the human process of thinking through oral assessments, reflective portfolios, iterative drafting with documented revisions, and real-time classroom critiques.
AI may help construct the digital skeleton, but students must still prove they own the heartbeat and flesh of the knowledge.
As Malaysia continues embracing AI across education, our assessment practices must evolve alongside it. If we can do that, we may finally stop policing whether AI is cheating our system and start asking a far more vital question: are we designing assessments that are still worth doing?
* The author is a Senior Lecturer at the Department of Language and Literacy Education, Faculty of Education, Universiti Malaya, and can be reached at [email protected]
** This is the personal opinion of the writer or publication and does not necessarily represent the views of NewsPulse.
