In A Pedagogy for Liberation (1987), Ira Shor and Paulo Freire criticised the banking model of education, in which learners are passive. In this model, learners are viewed as empty containers, and teachers deposit knowledge into them. In this process, education becomes an act of information transmission rather than an act of inquiry. They advocated for maintaining a dialogic relation with knowledge to develop critical understanding. AI’s over- and uncritical use reinforces the banking model, encouraging learners to rely on summaries, leaving no scope for deep engagement with the primary text. Concerns are more profound as AI enters the learning ecosystem. When AI is used in research, particularly in the social sciences, it changes how knowledge is consumed, produced, and evaluated. It is used at different stages of the research process, such as information search, literature review, data analysis, and writing papers (Patterson, 2025). While most journal publishers allow the use and disclosure of AI in research for grammar improvement, its use for other purposes remains a grey area.

Literature reviews conducted through AI lead to expertise without understanding (Morris, 2024). When a paper is read word-for-word, the researcher gains several additional details about the research area that may be relevant elsewhere or enhance understanding. Further, crucial debates may be missed when AI is used for this purpose. Rigorous research requires a deeper engagement with the journal articles that results from slow reading, re-reading, and reflection. This is not possible as of now with AI, as it falters with abstract concepts, and social sciences are replete with them.

One of the major concerns is that the AI-generated text remains a black box. Output varies each time, even when the prompt remains the same. Nonetheless, there is a perception that if the use of AI leads to greater productivity, the demand to be seen as productive will also rise as the new normal, affecting the well-being of researchers already in a publish-or-perish ecosystem (Chubb et al., 2022). Researchers may be tempted to over-rely on AI to produce more research papers at the expense of quality. Therefore, this process might reduce the quality of research papers published. Further, AI has overstrained the peer-review system of journals because more AI-assisted papers are now submitted (reducing the acceptance rate of journals and

Another valid concern is that over-reliance on AI hampers the skill development of research scholars. For instance, finding key papers, conducting a literature review to understand a research area, identifying research gaps, thinking and writing, in addition to data analysis skills, should be viewed as skills. From the Freirean perspective, AI’s usage in research may make future researchers passive consumers and producers of knowledge, undermining the goal of academic inquiry. Its use weakens critical thinking and argument-building skills, limits academic writing practice (a core skill set for a researcher) and engagement with the original text. Of these, writing is vital to being a good researcher. When new to research, researchers treat thinking and writing as two separate processes. However, writing is a part of thinking (Wellington, 2010) and fosters deeper understanding.

Further, there are broader issues of justice. Viewing AI from a Freirean perspective raises critical questions of justice, such as who benefits from it. AI exacerbates already existing research inequalities because scholars from the Global South lack access to advanced AI tools. Even when AI is used, it should be used to enhance the researcher’s critical agency. Another major issue is that AI systems are trained on large amounts of data, but the intellectual contributions of authors behind the data remain unacknowledged.

It is clear that human oversight is needed when AI is used in research because it may hallucinate information, cannot access articles behind paywalls (and thus may generate text based on biased, trained data), and risks the underdevelopment of research skills. While AI speeds up the research process, asking the right questions becomes more valuable for uncovering the dominant ideology of AI algorithms in the present times, as Freire would emphasise.

References

Chubb, J., Cowling, P., & Reed, D. (2022). Speeding up to keep up: Exploring the use of AI in the research process. AI & Society, 37(4), 1439–1457.

Morris, D. (2024). Magical thinking and the test of humanity: We have seen the danger of AI, and it is us. AI & Society, 39(6), 3047–3049.

Patterson, B. (2025). Can AI help with that? The limitations of AI tools for information discovery, search and reviews. Journal of Electronic Resources in Medical Libraries, 22(1–2), 56–59.

Shor, I., & Freire, P. (1987). A pedagogy for liberation: Dialogues on transforming education. Bergin & Garvey.

Wellington, J. (2010). More than a matter of cognition: An exploration of affective writing problems of post-graduate students and their possible solutions. Teaching in Higher Education, 15(2), 135–150.

***

Amit Yadav is a research scholar at the School of Humanities and Social Sciences, National Institute of Science Education and Research (NISER), Bhubaneswar, Homi Bhabha National Institute (HBNI), Training School Complex, Anushaktinagar, Mumbai.