Artificial Intelligence and literature reviews: can AI really improve scientific research?

As artificial intelligence is gradually becoming embedded in research practices, one question remains central: can it truly help improve the quality of scientific literature reviews? Drawing on the work of Prof. Xavier Castañer (HEC Lausanne, University of Lausanne) and Prof. Nuno Oliveira (Tilburg University), this article explores the role of AI in knowledge production, highlighting its contributions, its limitations, and the theoretical and ethical requirements that accompany its use.

Artificial intelligence now occupies an increasingly prominent place in academic work. In particular, tools capable of analyzing vast corpora of publications, mapping research fields, or identifying conceptual regularities have generated strong interest for literature reviews. Their promise is clear: to support the exploration of an ever-growing body of scientific knowledge. But what can AI truly contribute to an exercise as central and as demanding as the literature review?

In the sciences, including management and organizational studies, two concepts play a key role in assessing scientific quality: parsimony the ability to explain complex phenomena with a limited number of clearly defined concepts—and cumulativeness, understood as the capacity of research to integrate coherently into an existing body of work. Despite their importance, these principles remain difficult to achieve. The literature reports recurrent phenomena such as conceptual redundancy or the difficulty of linking findings across successive studies. These challenges are structural and long predate the recent rise of AI.

It is precisely within this context that the reflection developed by Xavier Castañer and his co-author is situated. Their work emphasizes that AI can play a valuable role as an accelerator of scientific work. In particular, it enables the identification of research streams, the synthesis of large volumes of publications, and the support of large-scale bibliometric analysis. When used rigorously, it thus provides significant methodological support for literature reviews.

However, AI neither replaces theoretical judgment nor the work of conceptual construction. It identifies patterns, but does not determine their scientific relevance. It does not rank concepts, arbitrate between competing theoretical frameworks, or produce, on its own, the meaning required for the advancement of knowledge. As Xavier Castañer and Nuno Oliveira emphasize: “Artificial intelligence can help researchers navigate the literature, but it cannot substitute for the work of meaning-making, theoretical clarification, and conceptual coherence that remains at the core of the researcher’s craft.”

The literature review thus remains a fundamentally human intellectual endeavor. It is based on explicit choices, the ability to connect works with one another, and an ethical responsibility regarding the use of the tools employed. This also entails heightened vigilance toward potential biases embedded in the corpora analyzed or in the models used.

The authors’ conclusion is unambiguous: AI is not a shortcut to better science, but a tool. It will not, on its own, resolve the persistent problems of parsimony and cumulativeness in management research. However, when used in an informed, ethical, and theoretically demanding manner, it offers considerable potential to support researchers in an ever-expanding scientific landscape. The future of literature-based research therefore rests on a productive balance between human intelligence and computational power, in the service of a clearer, more cumulative, and more robust science.

Reference :

Castañer, X., & Oliveira, N. (2020). Collaboration, coordination, and cooperation among
organizations: Establishing the distinctive meanings of these terms through a systematic
literature review. Journal of Management, 46(6), 965-1001.