In February, sections of the media reported that the University Grants Commission had returned several PhD theses submitted to the Babasaheb Bhimrao Ambedkar Bihar University in Muzaffarnagar. The commission found that around 40% of the theses had been written using artificial intelligence. The university was instructed to ask the researchers to rewrite and resubmit their work.
Despite this, the commission has not yet issued any formal guidelines on the use of AI in academic research. In June, The Times of India reported that the UGC had issued revised guidelines stating that a PhD with 10% to 40% of content generated by AI or which has been plagiarised will be returned to the scholar for revision within six months.
AI would be permitted only for language correction and other limited support, the article said. These guidelines did not appear to be immediately available on the UGC website.
But the guidelines, as reported by the newspaper, raise an important question: can the use of AI in research be treated as plagiarism?
Good research is expected to contribute new knowledge to the existing body of scholarship and open up new avenues for further research. To achieve this, a researcher must perform a few basic tasks.
First, they must familiarise themselves with the existing scholarship in the chosen field and identify what is often described as a “research gap”. This involves reviewing earlier books, articles, and research papers – an exercise commonly known as the “literature review”.
Once this stage is complete, the researcher must formulate a hypothesis or thesis statement and clearly define the objectives of the study. In practice, researchers do not perform all these tasks on their own. The guidance of a research supervisor is an integral part of the process.
AI tools can assist with certain aspects of this intellectual labour – suggesting directions, helping to organise ideas, or assisting in the search for relevant works to write the literature review. Should such assistance be regarded as improper?
That said, a note of caution is necessary. AI tools sometimes cite books and articles that do not exist in reality – a phenomenon known as “hallucination”. They may also generate hypotheses that are impossible to test empirically. A literature review produced entirely through AI, therefore, carries a high risk of error.
Similarly, the feasibility of a hypothesis or the practicality of research objectives must always be discussed with and validated by the supervisor.
In such cases, AI should function only as an auxiliary aid.
Research also demands extensive reading. Scholars must consult large numbers of articles, books, and research papers to know what they should exactly do. However, not all texts require the same level of attention. As Francis Bacon famously observed in his essay Of Studies, some books are to be tasted, others swallowed, and a few chewed and digested.
Likewise, a researcher reads some works closely while merely skimming others. For texts whose essential argument can be grasped through a summary, many scholars today use AI tools for having summaries of those texts. Should this also be deemed unacceptable?
In scientific research, AI-powered search tools can also rapidly identify relevant information and references. In certain disciplines, AI may even form part of the methodological apparatus of research. Fields such as information technology, data science, and digital humanities can employ machine learning techniques to analyse large datasets.
However, even in such cases, researchers must possess a clear understanding of the methodology and must interpret the results themselves, in consultation with their supervisors.
Already, several foreign universities have already issued detailed directives on how to appropriately use AI in academic research.
For instance, Oxford university has clearly specified the permissible areas for AI use in research. These include language refinement and correction of grammatical errors in theses, organising research sources and generating references but requires transparency, accountability and documentation.
In India, however, institutions such as Calcutta University and Shiv Nagar University have introduced regulations whereby researchers are not permitted to submit their theses if the use of artificial intelligence is found to exceed 10%.
While taking decisions on AI use in PhD theses, it should also be remembered that the software tools designed to detect the use of artificial intelligence in research are not always completely reliable.
The reality is that, regardless of the preferences of the University Grants Commission, researchers will inevitably use artificial intelligence in future academic work. Therefore, it is imperative to raise awareness among researchers about the proper use of AI.
In India, coursework is now mandatory at the beginning of a PhD. During this phase, it should be ensured that researchers attain AI literacy.
However, it is important to remember that even if AI is used in research, the true strength of a PhD thesis lies in the researcher’s original thinking, analytical ability and creativity. Thus, while AI may be employed as an assistive tool, the core intellectual work of research must be carried out by the researchers themselves.
While it is futile to ban AI in research, it is essential for scholars to understand the potential and limits of AI as an auxiliary tool in research.
Angshuman Kar is a professor in the department of English and culture studies at the University of Burdwan.
