AI and Thinking: Erosion or Enhancement?*

*AI and Thinking: Erosion or Enhancement?*

Author: Professor Tan Eng Chye

*The author is the President of the National University of Singapore*

Translated by ChatGPT

For subscribers only

https://www.zaobao.com.sg/forum/views/story20260520-9080662?utm_source=android-share&utm_medium=app

《联合早报》

Lianhe Zaobao 

2026-05-20

=====

Human learning begins with asking questions. Yet AI has dissolved the necessity of independent inquiry, sparing us the intellectual journey of repeatedly searching, probing, and refining meanings layer by layer until clarity finally emerges.

Walk into any seminar room at Kent Ridge, and you can sense that the foundations are quietly shifting. This is not because the desire for knowledge is fading, nor because the flame of curiosity is dimming, but because the very nature of “education” itself is being fundamentally reshaped. In the past, learning resembled climbing a mountain: searching through references, debating, getting lost, taking detours, reorienting oneself, and continuing the ascent before finally earning that expansive summit view. Today, information arrives instantly, packaged in fluent paragraphs, clear bullet points, and step-by-step solutions. Reaching the summit now takes only moments, while the arduous journey itself has somehow vanished.

AI has clearly accelerated learning. But has it also sharpened our critical thinking? Or is this precious human faculty now at risk of atrophy?

*Concerns Over Capability Degeneration in the AI Era*

Education has always rested upon two pillars: learning and thinking. Learning refers to absorbing new facts, frameworks, terminology, methods, and models of thought—in short, acquiring knowledge. Yet the acquisition of knowledge alone is insufficient. Only when critical thinking enters the process does learning become complete: when we begin questioning, dissecting, and interrogating what we have learned, and from there derive new conclusions, viewpoints, and theories.

Today, these two pillars are no longer advancing in tandem. AI has taken over the laborious work of gathering information and transmitting knowledge, but it does not necessarily help us cultivate the most essential ability of all: critical thinking. One could even say it is quietly pushing thinking toward decline. Many of us are falling into the risk of treating AI as a “GPS for thought.” It undoubtedly provides the convenience of immediate answers, but it also dulls a more fundamental capability: the ability to draw our own “mental maps” needed to navigate complex problems when no technological guidance is available. Once we only know how to follow prompts, we eventually lose our sense of direction.

The real world has already validated these concerns.

In 2021, a study observed changes in the skills of endoscopists before and after the introduction of an AI tool designed to identify precancerous lesions during colonoscopy procedures. The results showed that three months after the AI tool was introduced, when doctors performed colonoscopies independently without AI assistance, the detection rate of precancerous lesions fell by 6 percent. This suggests that the use of AI in diagnostic colonoscopy indeed carries the risk of skill degradation among physicians.

Similarly, last year, researchers at the MIT Media Lab found that subjects who used generative AI to write essays displayed a significantly weaker sense of ownership over their work. Electroencephalogram (EEG) results—which measure electrical activity in the brain—also revealed weaker neural connectivity. In other words, fewer regions of the brain were engaged collaboratively during the writing process. Over time, these subjects continued to show declines across neurological, linguistic, and behavioral dimensions, once again raising concerns about skill degradation.

The risks brought about by AI go beyond mere “de-skilling.” Another equally subtle danger lies hidden beneath the surface: “never-skilling,” where individuals lose the opportunity to develop foundational abilities from the very beginning. Human learning begins with asking questions, whether through the endless “whys” of children or the Socratic method that remains widely used in classrooms today. Yet AI dissolves the necessity of independent questioning and removes the intellectual journey of repeatedly exploring and refining meaning until understanding finally dawns. Paid premium AI systems intensify this effect: they are more capable, and their answers and guidance appear more polished and rigorous. Ironically, this heightened appearance of reliability gives rise to automation bias. Even when contradictions exist within the information, students often absorb entire passages without question, gradually losing the instinct for critical thought.

*Taking Back the Steering Wheel*

AI is now an established reality. What we need are not walls and barriers, but practical guardrails. One example is the “AI-free periods” implemented by the medical organization at the National University of Singapore, during which healthcare professionals are prohibited from using any AI tools in clinical work or medical evaluations. In the financial sector, the Monetary Authority of Singapore is formulating guidelines on AI risk management aimed at promoting the safe and responsible use of AI. Financial institutions are required to ensure that AI serves an assistive role rather than replacing human judgment in high-risk decision-making scenarios. Before final decisions are made, professionals must review and question AI-generated outputs. The purpose is to prevent the erosion of professional expertise caused by overreliance on technology.

This does not mean rejecting AI. In fact, from the NUS medical organization to major healthcare systems worldwide, AI has already been widely applied in assisting diagnoses, optimizing treatment plans, and managing medical cases. The key lies in striking the right balance between the two.

The aviation industry offers a particularly illuminating reference point. During cruising, autopilot systems are commonly engaged, allowing pilots to focus on monitoring, observation, and communication with air traffic control. But during the two critical phases of takeoff and landing, autopilot rarely intervenes, and pilots must operate the aircraft themselves. Aviation regulations further stipulate that pilots must complete a required number of takeoffs and landings in order to remain qualified to fly, ensuring their skills remain sharp, their instincts intact, and their control over the aircraft uncompromised. Likewise, we must carefully navigate the delicate boundary between automation and human control. We must continue familiarizing ourselves with AI and using it effectively, while also insisting on training and practicing without AI assistance. Only then can we preserve our fundamental competencies.

So what does all this mean for educators today? AI poses a fundamental challenge to pedagogy. We must completely rethink how teaching is conducted, yet the fundamental direction of education remains unchanged: strengthening the foundations of thought. AI should not replace the fundamentals of learning and thinking, just as calculators never made mental arithmetic obsolete. Ultimately, pedagogy must guide technology—not the other way around, where technology dictates pedagogy.

This is precisely why some institutions are returning to basics: supervised pen-and-paper examinations, oral defenses, and restrictions on internet usage during key assessments. This withdrawal from the “take-home assignment” model is not a rejection of technology, but rather a safeguard to ensure that genuine learning occurs and that core competencies do not quietly decay. Nor will this challenge end at the campus gates. Once students enter the workforce, we must continue maintaining and honing these abilities throughout their professional lives, ensuring that hard-earned expertise and critical judgment do not dull and fade over time.

More importantly, beyond cultivating users who can skillfully and confidently operate AI, we must also nurture thinkers capable of leading technology itself. This means cultivating people who truly understand how AI systems are built, how they are controlled, and—most crucially—how they should be examined and questioned.

Whether climate scientists use AI to accelerate breakthroughs in global medium-range weather forecasting, legal professionals employ AI to deepen legal research and case analysis, or urban planners model sustainable cities of the future with AI, they are not merely users of technology. They are AI architects within their respective fields—both shapers and guardians of technology. Firmly holding the steering wheel, they guide AI with sharp insight, responding to the world’s challenges and solving the defining problems of their era within their own disciplines.

*In the AI Era, Humanity Becomes More Distinct*

Paradoxically, AI helps us better recognize our own foundations and authentic nature: it is precisely the existence of the “artificial” that makes the contours of humanity stand out more clearly. Complex reasoning, deep analysis, ethical judgment, empathy, and warm human connection—these capacities together form the essence of what makes us human, and they are precisely what we should devote ourselves to cultivating. If used properly, AI will not replace humanity; instead, it will amplify it. Through this, we may grow, flourish, and become more fully human.

The road ahead will inevitably include failed attempts, lessons learned, moments of self-reflection, and flashes of sudden clarity. If we remain ambitious and refuse to settle for the easy path, the heights we may reach will be astonishing. In the end, the true measure of success has never been how sophisticated our tools are, but whether we can cultivate resilience and creativity within our teams. If our direction is correct, we will not merely respond passively to the future—we will shape it.

The author is the President of the National University of Singapore.