From Scarcity to Abundance: Scholars and Scholarship in an Age of Generative Artificial Intelligence

Abstract

Generative artificial intelligence (AI) refers to a class of machine learning technologies that have the capability to generate new content that resembles human-created output, such as images, text, audio, and videos.Within the context of generative AI, much hype has been directed in particular toward large language models, given the associated tools' ability to generate coherent and contextually relevant text given a user-defined prompt.As editors of a journal devoted to advancing knowledge production in the areas of management and organizations, we believe it is important to recognize the promise of generative AI for advancing such knowledge production.In particular, we see opportunities for scholars to increasingly make use of generative AI to assist with the entire value chain of knowledge production, from synthesis, to creation, to evaluation and translation (Bartunek, Rynes, & Daft, 2001;Kilduff, Mehra, & Dunn, 2011;Van De Ven & Johnson, 2006).Moreover, these same tools promise to increase both the efficiency and rigor of our research methods.Even as we recognize the promise of applying this technology to the craft of scholarly knowledge production, we also recognize the potential perils of doing so.The integrity of the scholarly research process depends upon methodological transparency and reliability.Yet implementation of large language models at scale poses significant risks to such transparency and reliability, even as it introduces additional concerns regarding accuracy, such as "AI hallucinations" and what we are calling "deep research fakes" (Walters & Wilder, 2023).Where the former refers to coherent, contextually relevant, yet false information provided by generative AI that may deceive the researcher, the latter refers to the intentional use of generative AI to create and manipulate data in order to deceive the scholarly community.Moreover, as with all forms of craft, as technology begins to increasingly mediate the human-centered foundations of scholarship, this is likely to raise questions of "authentic" scholarship (Kroezen, Ravasi, Sasaki, _ Zebrowska, & Suddaby, 2021; Voronov, Foster, Patriotta, & Weber, 2023), and indeed under what conditions human judgment may be necessary to uphold such authenticity.Taken together, the promise and peril of generative AI raise a number of critical questions.First, how might generative AI be used to increase both the quality and quantity of interesting and important scholarship?Second, given the aforementioned tensions surrounding its usage, why should we caution the usage of generative AI?To date, prior editorials and publications from across the academic profession, including a recent Academy of Management Journal editorial (Dwivedi et al., 2023;Susarla, Gopal, Thatcher, & Sarker, 2023;Thorp, 2023;von Krogh, Roberson, & Gruber, 2023), have begun to consider both of these pragmatic and normative questions in light of both the underlying technology and methodology.However, given the broad and nontechnical appeal of generative AI through tools such as ChatGPT and Microsoft Copilot, we see an opportunity to advance this conversation toward a broader and more future-oriented investigation as to the implications of generative AI for the scholarly profession, and for journals like the Academy of Management Journal in particular.To be clear, as we address this third and final question, our intention is not to codify our profession's response to the future development of this technology, but rather to expose various possible uncertainties that will affect its development and its implications for scholars and scholarship.In other words, we are merely at the beginning of a conversation we expect to be having for many years to come. HOW MIGHT GENERATIVE AI BE USED?In order to understand how generative AI might be used to enhance the quality and quantity of scholarly knowledge production, it is worth deconstructing the scholarly process into its constitutive parts, which we suggest entail (a) knowledge synthesis, (b) knowledge development, (c) knowledge evaluation, and (d) knowledge translation.While some of these steps rely more heavily on "know-what"-the knowledge of data, information, concepts, and 1617

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 179 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.