In my work as a philosopher, writing is just as essential as thinking. In fact, writing has always been a way of forcing myself to put my thoughts and ideas into words, to see if they make any sense, whether my arguments are valid or not, and so on. Yet, with the introduction of generative artificial intelligence (GenAI), things have changed. As reported by the news site DailyNous, one paper published in the journal Philosophy and Public Affairs was largely written by the AI tool Claude, while the philosopher listed as the author “acted as a generous PhD supervisor for Claude.” Unsurprisingly, this has fueled a debate within academic philosophy. One question is whether certain AI tools, particularly Large Language Models (LLMs), could qualify as co-authors. In this blog post, I am less interested in this question. Rather, I would like to take the opportunity to reflect on different ideal types of academic authorship and their implications for academic writing broadly, and for the use of GenAI in particular.
It is common knowledge in the field of publication ethics that there are different traditions concerning authorship in different fields of research. In some fields, single authorship is still the norm, while most fields are largely cooperative. In some fields, group authorship exists, while in others, authorship is primarily viewed in terms of individual contributions rather than a group effort.
To highlight differences in disciplinary perspectives, I will make a distinction between two ideal types of academic authorship. These two ideals are by no means intended to be exclusionary or incompatible with one another. Neither are they intended as factual descriptions, nor are they here presented as admirable or desirable ends. Rather, they are idealizations in the Weberian sense. As such they are here used in order to structure an otherwise chaotic social order and allow me to focus the discussion about academic writing and LLMs in a way that I consider illuminating.
The first ideal I call the author as intellectually invested. By intellectually invested, I mean that the author is someone who
- sees their text as their own, that is, they see themselves as the creator and originator of the text, and
- cares about the text, how it is structured as well as the ideas that it seeks to convey, and who is ready to defend the text in terms of its form and content, and
- experiences a personal responsibility for eventual mistakes or mishaps in the text.
These conditions can be seen as both necessary and desirable characteristics of academic authorship. If one does not see the text as one’s own or does not care about the ideas it seeks to develop, one simply does not fulfil the criteria for intellectual engagement that follow from academic authorship.
I call the second ideal the author as a mere rapporteur. According to this ideal, academic authorship only involves the truthful and correct description of what has been done during the research process, that is, what has been done, why and how the research was performed and the results thereof. From the perspective of this ideal, the interesting part of the research has already taken place, and the authoring of the paper is simply a matter of writing a truthful report.
To be sure, these two ideals of authorship may overlap. There is no direct opposition or conflict between them. Someone who writes a truthful report may also feel a strong sense of personal responsibility for the text, and they might see themselves as the text’s originator. At the same time, I believe that there are good reasons to keep the two ideals apart, as they highlight very different aspects of academic authorship – aspects that may also be relevant to the varying degrees of acceptance regarding the use of LLMs in academic writing.
Those who primarily see the authorship of an academic text as a matter of truthfully reporting what happened in the research process have a largely instrumentalist view of the text. The value of the text is simply a matter of to what degree it represents the reality of what is reported and how well this is communicated to the wider academic community. In contrast, those who see academic authorship as a matter of being intellectually invested in the text will most likely view things differently. In their view, the text is an intellectual product and an expression of one’s own intellectual engagement with the issue being addressed. As such, the text is not purely instrumental. Among other things, it lends a sense of pride or personal growth in a way that one does not feel for other academic texts. The text has a personal value for the author beyond whether and to what extent it reflects the reality of what is being reported.
The two ideals of academic authorship may also have importance for how we approach the question of to what extent the use of GenAI should be encouraged in academic writing. For those who first and foremost see authorship in an instrumental way, it is not clear why they should be the ones who draft the paper. What they need to do is first and foremost to make sure that what is reported in the text is accurate in terms of how well it represents the research that has been conducted. One’s responsibility as “a writer” is about just that. How the text came about seems less relevant. Using GenAI to draft one’s research paper therefore seems to be perfectly legitimate insofar as there are no further external wrongs. However, if we believe that academic authorship also requires that the writer is intellectually invested, the use of GenAI becomes less attractive. Since this ideal of authorship assumes that the authors see themselves as the originators of the text, this seems to rule out the use of generative AI, perhaps already at the draft stage.
Does it matter which ideal of authorship we endorse? As already mentioned, it is widely noted that there are different authorship practices in different areas of research. Although there is no necessary commitment to a particular ideal within actual fields, the two ideals tie into different disciplinary practices. The author as a mere rapporteur aligns well with the practices of authorship that are found in the medical and natural sciences. In these fields, the research is what takes place in the lab, during field experiments, observations etc. Research articles of such fields also typically follow a specific structure and writing them is merely a matter of communicating the research that has already been done. In contrast, in a lot of work in the humanities, writing is part of the research process, and the author is very much intellectually invested in it. To write is to do research. The writing of the text is as much a part of the research as one’s results or conclusions. This, I think, might also explain why some academic fields, as well as individual academics, are less enthusiastic about GenAI.

Written by…
William Bülow O’Nils, senior lecturer in biomedical ethics at the Centre for Research Ethics & Bioethics.

0 Comments
1 Pingback