Mumbai/New Delhi: “Reckless use of AI” in setting and translating exam papers, followed by an unexplained six-week delay in cancelling three flawed tests, is at the heart of the UGC-NET controversy that’s forced National Testing Agency to order fresh exams in English, commerce and sociology, according to subject experts and NTA insiders.The testing agency, however, denied generative AI was used to author the disputed papers.The experts told TOI on condition of anonymity that AI was used extensively to prepare the papers. Former subject experts associated with NTA till recently said AI had been deployed not only in setting UGC-NET questions but also in translating NEET-UG papers. They attributed the scale and nature of the errors to inadequate human scrutiny.The controversy has drawn attention to the extent to which NTA has integrated AI into the preparation of examination material, particularly translation. “We have had several meetings with two senior officers of the NTA who had confirmed that they were unable to declare the answer keys as the papers were designed in an ad hoc manner. When asked what ad hoc meant, they said nothing,” said a senior source in the education ministry. “Asked why the question papers were not reviewed by experts, there were really no answers,” added the source.The agency’s reliance on technology is not, in itself, evidence that AI caused the errors in the cancelled exams.Another source said, “A panel of four to five experts sets a question paper. If there was indeed a panel, how were so many questions from the previous year’s question papers picked? How did the examiners not have answers to the questions they had designed?”NTA director general Abhishek Singh, though, denied generative AI was used to author the disputed papers. He told TOI there was a “human review” after the papers were set. When asked how the errors went unnoticed then, he had nothing to add.Also under scrutiny is NTA’s response after the discrepancies surfaced. Education ministry officials and NTA sources said the ministry favoured cancelling the papers and start retesting by July-end or the first week of Aug. NTA, however, did not act on the advice, waiting nearly six weeks and announcing the cancellations only when it released provisional answer keys for 84 of 87 subjects.The sociology exam attracted strong criticism. Candidates complained of misspelt names of internationally known scholars, garbled terminology, awkward translations and questions, they said, bore little resemblance to the prescribed syllabus. George Ritzer reportedly appeared as “Putzer”, Talcott Parsons as “Parsow”, G S Ghurye as “Ghunye”, and Martha Nussbaum’s name was also distorted.The English paper raised a separate concern over large-scale repetition. Candidates alleged that around 67 of the 150 questions had appeared in an earlier UGC-NET exam, with some claiming that even the sequence of answer choices had been retained.Commerce candidates also reported that several questions had been repeated from earlier exams, including questions relating to taxation, break-even analysis, capital structure and services marketing. How those repetitions occurred has not been conclusively established.“The larger issue is the degree to which automated systems were used to translate, organise and assemble high-stakes national examinations and whether adequate human academic review took place after that process was complete,” said a faculty from JNU. UGC-NET determines eligibility for academic and research careers across Indian universities, making the reliability of the exam process especially significant.
Share your thoughts in the comments
Be respectful · TOI community guidelines
The fresh claims deepen the controversy beyond poor paper-setting. They raise questions about who authorised the use of AI, which tools were deployed, whether qualified experts vetted every question and why NTA delayed cancellation despite the ministry’s view.The central question is no longer simply how so many mistakes entered three national-level examinations. It is how much of the exam process had been handed to machines and whether enough qualified humans were checking what the machines produced, experts said. For an exam determining eligibility for teaching and research careers, the issue is whether automation was allowed to replace academic judgement and accountability.
