Expert tells why Western AI models are better for education

PETALING JAYA: Western artificial intelligence engines are a better option for those who want to use AI for education and research purposes, particularly compared with platforms from China, says an IT expert.

SL Rajesh of the International Association for Counterterrorism and Security Professionals Centre said one reason was the “language trap” that Chinese AI faces.

He said this is where an AI may lack the linguistic strength to translate reasoning, nuance and knowledge, despite being “fluent” in communicating with users in multiple languages.

“For example, an AI may answer perfectly in Chinese but struggle with complex academic writings in English. It may translate words correctly but miss cultural contexts, idioms, or subtle meanings.

“Fluency can make an AI appear equally capable in two languages, even when its quality of reasoning isn’t actually equal across both. For education, that difference can affect understanding, essay writing, and critical thinking,” he told FMT.

On the other hand, Rajesh said, Western AI models are trained on very large amounts of multilingual data from academic, scientific, and public sources, much of which are published in English.

“As a result, they often handle English academic writing more naturally, understand international textbooks and research papers better, and switch between languages while keeping the same level of reasoning,” he said.

He pointed out that AI firms like OpenAI, Google DeepMind, and Anthropic have strong ties to international research communities, allowing researchers to evaluate the AI model’s capabilities and limitations easily while giving feedback accordingly.

Their models are also widely used by universities, publishers, and educational technology providers worldwide, he said.

Rajesh said another issue was that Chinese platforms operate under strict content regulation by Beijing, affecting the topics that users may research.

One widely reported example is how platforms like DeepSeek self-censor when it comes to queries about the 1989 Tiananmen Square massacre or the Umbrella Movement protests in Hong Kong, with Chinese AI systems replying that these topics fall beyond their scope.

“Some Chinese AI systems may avoid or restrict certain sensitive topics, which can limit deeper academic discussion. In higher education, students are expected to debate and explore all sides of an issue,” he said.

China excelling in some areas, but more openness needed

Rajesh pointed out that Chinese AI has made great strides in some areas, especially in mathematics and coding. With this in mind, he said, the “best” platform was partly subjective to the field of study.

Chinese systems would be able to catch up with the West if they improved their English and multilingual performance, become more open when it comes to topics, and focus on developing AI that helps users think critically, he said.

He also said they should collaborate more with the world’s top universities and use more international research in its data training, though this goes back to being open to sensitive topics and critique.

Humans equally susceptible to biases

SP Logendra, head of Taylor’s University’s school of education, said biases were not limited to AI engines as humans are also susceptible to cognitive biases and flawed narratives.

Logendra said the authors of history textbooks would have also had their own worldviews, while the general public rarely analyses this or whether the information contained was grounded in facts or curated narratives.

With this in mind, he said, it was crucial for educational institutions to pivot towards “explicit AI literacy”, where students are encouraged to use a variety of platforms instead of just one.

At the end of the day, the effectiveness of AI engines as education tools ultimately lay in the hands of their user, Logendra said.

“We should train them to treat every single AI response as a subjective ‘working draft’ rather than an absolute authority.

“By teaching students to cross-reference multiple AI models, verify source citations, and actively look for missing historical or cultural perspectives, we turn the weaknesses of AI into a powerful exercise in critical thinking and media literacy,” he said.

Rajesh agreed, saying that students young and old must be trained to treat AI as a tool for learning rather than a source of truth. “Always verify, compare, and think critically,” he said.