Each one of us has questions throughout our life. And we know the answers to only a few. Efforts of varying degrees go into getting the answer. We can ask our parents, teachers, friends, a book, an article, the internet or an expert who helps us find the answer. The process takes time, which means many failed attempts and a lot of reflections. But since 2023, we have a new answering tool that knows it all- Artificial Intelligence. And this tool has become much smarter with lesser mistakes in 2026.
(Sign up for THEdge, The Hindu’s weekly education newsletter.)
“The unexamined life is not worth living”, the words supposedly uttered by Socrates, who was forced out of the world for just questioning, need a revival today. Repeated questioning and critical thinking have been the foundation of human intelligence for over 2,000 years. What if we encounter a situation when questioning itself becomes a huge effort for humans.
Let us take a scenario where a student shares a part of his lesson with the AI tool and asks it to explain it in an easy-to-understand way. This is a type of human-AI communication that we can call source-based use of artificial intelligence. In such interactions, any question asked by the student is answered by the AI tool considering the content given in the source. These answers can be easily verified with a teacher or a related reference book. The good thing is that such source-dependent use of AI, coupled with a prompt to the tool instructing it not to self-imagine (hallucinate), gives accurate response most of the time.
But for a student at a higher education level or a researcher whose learning is based on gathering a large amount of information that is not restricted to a few sources, the use of AI has different consequences. It is an undeniable fact that most, if not all, researchers use AI for various purposes. AI assistance can be in the form of reviewing a concept they have developed on their own, suggestions to overcome a laboratory hurdle they are facing or even to develop a research question from scratch.
In such a situation humans are faced with the information generated by the AI tool which is trained with almost all available resources on the topic searched. The natural belief of the learner is that the accuracy of the response is likely to be fairly correct. When you can have an accurate answer to a complex question, our mind will think, “Why should I confirm the accuracy of the response by spending long hours following the traditional way?” This is exactly the thought which leads to deeper passive conversation, barring a few questions between the learner and the AI . One question the learner forgets to ask is , “Should I challenge the AI response with my own knowledge?”
Instead the learner asks, “Can I ever be able to challenge the highly knowledgeable AI?” Most of us accept defeat right in the beginning. Then where is the game?
Attaining high level of understanding
When a learner decides to use AI assistance and really wants to develop a high level of understanding, the first effort should be to learn all the basic knowledge relevant to the subject through authentic books, research articles, experts’ ideas, and other conventional resources. This gives the basic skill required by the learner to guide the AI tool during conversation. Lack of such basic knowledge will take the AI down a random path and gives a bigger opportunity for the AI to speculate, imagine and mislead the learner to a less resourceful end.
That said, even when the researcher has prepared well before approaching the AI tool, there will be many ideas thrown up by the AI tool, which appear to be innovative but new to the learner. Here, the learner may be tempted to proceed further with the innovative suggestion without examining the idea more deeply through traditional literature.
Lack of time may be the reason users say for this. But we should remember that AI misses a large proportion of articles on any topic, frequently even the best ones. At every step of the AI conversation, the learner will face the challenge of challenging the AI’s ideas. At times, even experts in the field can be taken aback when they fail to understand what the AI model is suggesting. The innovative information may appear groundbreaking at first look. But only a deeper verification will help confirm or discard the idea.
When one tries to challenge AI at every level, time will be lost. But gain occurs in deep learning of the content, which would not have occurred if not for the AI that popped up the suggestion in the first place. If the AI realizes that the user is challenging what it says, the neural network path changes towards more advanced thinking. When the users’ responses defeat or revise the AI’s thought, subsequent responses become more cautious. This leads to a more refined output from the AI tool.
Most of us think that the use of AI will decrease the time taken to arrive at novel research ideas. Effective AI use will, in fact, lead to more time for developing innovative concepts, but their quality will be at a much higher level. The quality can be as good as an idea developed with the help of a pioneer on the subject, affiliated with a premier institution who is not accessible to most. AI has the capability to guide the user to that level. Provided you challenge it intellectually.
There is another side to the challenge, which is “Self and society”. Humans tend to go with the tide. When most of their peers are moving fast in their research with passive use of AI, one’s mind will hesitate to take the road less travelled. Here, educationalists and research guides should constantly instruct their students on the benefits of time-tested, slow and steady learning of concepts, which may need hours of reading, visits to real laboratories or places, and reflection on the learnt content. What they should not do is passively accept it or replicate the same mistakes which their students make.
Emerging challenges
Let me close with two disturbing situations when AI use can challenge the entire scientific community and create an environment of mistrust and inappropriate judgement.
An emerging challenge is when invited peer reviewers of research articles use AI to review original research. These reviewers have the liberty to ask questions about the research without a reference and can quote it as their own experience. This turns into a burden for the article’s author, who is left to defend the questions, which may look innovative but may not have a strong current scientific basis or may be just cunningly speculative.
Public viva-voce is another place where AI generated questions can cause havoc to the presenter, especially when asked by an expert on the subject. We tend to believe that an impressive question asked by the subject expert deserves justification on the floor. But that is impossible since the question itself has no plausible answer.
We may not be able to do much on this through rules and regulation. Self-discipline and scientific integrity can. When researchers use AI to draft methodologies through passive, poorly explored AI conversations, they write articles that can succeed in the basic scrutiny for a degree but arrive at conclusions which may be half-truths. Personal objectives are attained. Scientific goals are buried.
Human intelligence has reached a stage where we can further improve many fold by challenging the new stranger or go for a free fall to thousands of years behind.
Dr. M. Emmanuel Bhaskar is a Professor of Medicine at the Sri Ramachandra Medical College and Research Institute, Chennai
Published – September 02, 2026 08:30 am IST
