The summer submission season is fast approaching. Some journals have jumped the gun and are opening well in advance of August 1, throwing professors' summer plans into disarray. The Virginia Law Review is one such innovator/offender.
Notably, the Virginia Law Review is requiring that authors disclose authors' "substantive" use of AI in the drafting process. They elaborate:
By “substantive,” we mean the use of AI to support an author’s factual assertions, legal claims, or overall argument, such as generating relevant cases, writing or revising textual sentences, or drafting parentheticals. The use of grammar or spelling assistance services such as Grammarly or spell-checkers is not considered “substantive.”
The law review "emphasize[s] that the use of AI is not disqualifying," but urges authors to "err on the side of disclosure." The law review reserves the right to rescind publication offers if they discover undisclosed AI usage.
In my unscientific observations of reactions to my post noting this new requirement and others' posts, I see a few concerns bubbling up. Some are (rightfully) worried that if many journals take a similar approach, it may become confusing and difficult to fill out all these disclosure forms. Virginia's, for example, was included in the submission guidelines that one can click to see on Scholastica, but it required an additional step by submitting authors to see it was there and it may be missed in the high-volume law review submission game.
Others seem worried over what level of detail detail is required in the disclosures, though the guidelines go a long way to specify what uses do not count and provide sample disclosures for authors to reference (and to emphasize that, when in doubt, more should be disclosed).
But the concern that got my attention is that authors will simply circumvent the disclosure by dishonestly claiming they did not use artificial intelligence to assist in the research or drafting of their scholarship. The disclosure is a simple Google form, and it's easy enough to state (as I did) that the author did not use AI in the process of researching and writing their scholarship (as I did not--more on why I did not here). The objection seems to be that authors could use AI to write their scholarship, make sure they've covered their tracks in the final work product, and falsely claim that no AI was used. The disclosure requirement, as a result, makes no difference.
I find this worry interesting for several reasons.
First, it seems to be at odds with a presumption that one's scholarly interlocutors are acting in good faith when advancing their claims and arguments. There have been plenty of accusations of bad faith flying around lately--particularly in the birthright citizenship space. Amidst the heated rhetoric, there have been occasional claims that the other side is acting in bad faith. They don't REALLY think that the Constitution supports their claims and are, instead, making arguments to get attention, judgeships, or sweet, sweet, Supreme Court citations. In response, those accused (and others beyond the debate) tend to object to these accusations of bad faith. The assumption that one's scholarly opponent is arguing in good faith is a crucial norm of the profession. If a substantive debate is to be had, one must treat one's opponent's arguments charitably. Accusations of bad faith circumvent genuine discourse and turn the endeavor into a shouting match.
I won't express any thoughts on the substance of this good faith debate (although I certainly have THOUGHTS). But this call to respect the motives of ones intellectual opponents isn't new, and underlies longstanding calls for charity when responding to scholarship with which one disagrees.
If scholars are operating under such a strong presumption of good faith, it puzzles me why so many assume that AI disclosure measures will be so easily and widely circumvented. Doing so involves lying to the editors. The black-and-white nature of the falsehood these critics foresee is arguably more profound than advancing arguments in bad faith--at least those arguments may still be colorable or have something in the way of merits (again, no opinions on those arguments expressed here; despite my THOUGHTS).
It seems like the longstanding presumption of good faith and charity disappears when we start to talk about AI disclosures. This raises questions over how seriously we take this presumption and whether its applications are limited. Perhaps we should assume that scholars are entirely honest in the arguments they (or their AI companions) put forth, but they're a pack of lying scoundrels when it comes to communicating with law students. I'd like to hope this is not true.
Second, expressions of concern over AI disclosure policies seem to assume there is some level of shame in using AI--or at least a perception that one will be shamed should their AI usage come to light. Avoiding shame explains why people might be willing to lie to law review editors. Folks typically don't lie without a reason. That reason might simply be the love of the game--which I respect in some contexts. But here, the motivation to circumvent disclosures seems to be the worry that one's work will be thought of as lesser because one used AI to write it.
I think this concern is perfectly reasonable. It is, after all, consistent with my own instincts. If I am given a paper to read, and I am told that AI was used to write it, I will likely question whether reading the paper is worth my time. After all, it wasn't worth the time of the author to write.
(There are of course, rejoinders that the substance may be fine. While this back-and-forth isn't the point of this post, I'll briefly note that these rejoinders speak past my objections: my distaste is based in the disrespect the person presenting me with the piece is showing me by expecting me to take time on something a machine spit out for them. The substance is immaterial to this breach of etiquette.)
To the extent that critics assail the Virginia Law Review's demand for disclosures out of a fear of this shame, there is a glimmer of hope that AI usage might not catch on as readily as its advocates suggest. A norm of respect for each other's time and insight is central to a community of scholarship. It wasn't easy for me to make it into the legal academy, and I confess I feel dismayed to see would-be colleagues turning their time and attention over to tinker with robots and glorify their outputs rather than engage in genuine conversation and debate. If people are ashamed of AI usage, though, it suggests that people recognize the loss of community and respect that occurs when genuine work is replaced by efficiently generated output. I'm not at all certain that the fear of this loss will keep everything from devolving into a dystopia where everyone uses AI to summarize everyone else's AI-generated output. But it gives me some hope.
Third, as much as critics claim that the Virginia Law Review's policy is easy to circumvent, I don't think this is entirely correct--at least for most authors. Even if we're willing to throw presumptions of good faith out the window when debating AI disclosure policies, circumventing these policies requires lying to law review editors. Lying isn't easy for most people. Those who seek to misrepresent their level of AI use must write out and submit falsehoods to the time-pressured, overworked student editors who will be reviewing (and maybe even editing!) the submitted article. False statements may be physically easy to write and submit, but I expect that doing so over and over again may begin to take a mental toll.
Perhaps fear of this mental toll will deter would-be wrongdoers from dishonest disclosures. More pessimistically, it may at least cause them some discomfort and heartburn, which is a bit of comfort for those of us who are dissatisfied with the dystopian, automated world that seems to be forming. The true psychopaths among us will lie away with abandon, but what else is new?
Maybe I'm wrong. Maybe the bad habits we've developed over years of playing the submissions game have caught up to us. Authors submit articles to journals in which they know they'll never publish--using the resulting acceptances as a means to an expedite. Editors respond with form rejection letters that contain no engagement or explanation of the decision. After years of enduring and perpetuating these little moments of disrespect, maybe it makes sense to expect that authors will be willing to pick up one more nasty habit in the interest of getting ahead.
I'm not sure how to respond to such a pessimistic vision other than to advocate for a different way of doing things and to act accordingly. I'm therefore in favor of Virginia's policy requiring disclosures, and I hope more law reviews take a similar approach. I'd love to see these forms integrated into Scholastica (perhaps in a manner similar to how Scholastica automatically requires anonymized submissions for journals with anonymous submission policies). I'll fill these forms out with ease as I continue to expend exponentially increasing effort to avoid generative AI in all its pernicious forms. And I'll hold out hope that others will do the same.
No comments:
Post a Comment