Hong Kong Educators Balance AI Integration With Cognitive Risks

As generative artificial intelligence becomes a staple in classrooms across Hong Kong, educators and researchers are grappling with a significant challenge: integrating the technology without undermining the fundamental development of student critical thinking and learning skills. While schools are rushing to adopt AI, recent data highlights the potential for serious academic setbacks if reliance on these tools goes unchecked.
A study conducted by researchers at Stockholm University and the University of Hong Kong (HKU) has raised concerns regarding the long-term impact of AI on academic performance. By tracking over 26,000 middle and high school students over 30 months, researchers observed that while AI use reduced homework time and initially boosted homework scores by 18 per cent, students’ monthly exam results plummeted by 20 per cent within six months. Over a two-year period, scores in high-stakes entrance examinations for high school and university fell by 24 per cent and 18 per cent, respectively. Researchers concluded that the technology is having a “substantial negative impact on student learning.”

Dragan Gasevic, chair professor of learning analytics and artificial intelligence at HKU, attributes this decline to “metacognitive laziness.” He warns that students may offload complex tasks to AI, producing work beyond their actual capabilities without internalizing the necessary knowledge. Furthermore, he notes that students risk losing the ability to distinguish their own understanding from information generated by AI.
Despite these risks, total bans on AI are viewed as impractical by many in the academic community. Professor Hui Li of the Education University of Hong Kong (EdUHK), who points out that research into the long-term cognitive effects of generative AI is still in its infancy, argues for a “nuanced, developmentally informed approach.” He suggests that for older students, the priority should be teaching them to navigate AI responsibly, verify information, and recognize inherent biases, as these tools will likely remain part of their professional and academic futures.

Local institutions are experimenting with diverse frameworks to maintain academic integrity. At the International Montessori School (IMS), elementary students are taught AI awareness without direct access to chatbots, while middle schoolers utilize closed, teacher-monitored platforms designed to act as Socratic study companions. Similarly, Bloom KKCA Academy employs a “traffic-light” system to dictate the levels of AI interaction allowed for various assignments, utilizing AI primarily as a “thought partner” for research and reflection.
Some experts are focusing on redesigning assessment methods to ensure they remain “AI-resilient.” Professor Cecilia Chan of HKU developed the TACO framework—Think, Ask, Check, Own—to encourage students to maintain independence in their cognitive processes. Chan, who also created the SuperTA tool to help educators identify vulnerabilities in their assignments, warns against over-reliance on AI detection software, which she argues can turn teachers into “policemen.”

Ultimately, the goal for many schools is to ensure students remain what Ellen Jeffcott, head of teaching and curriculum for the secondary section at Yew Chung International School, calls the “thinker-in-chief.” The consensus among these educators is that while AI can assist in refining lines of inquiry and organizing information, it must not replace the student’s own capacity for evidence evaluation and argument formation.
While initiatives to teach AI literacy are growing, a report from the Our Hong Kong Foundation (OHKF) found that one in four local students still struggle to complete homework without the help of AI, and many students have already achieved a level of proficiency that outpaces their instructors, complicating efforts to monitor its use.