Chat

Chat

1

We have been scrambling to put together this issue of The Brief. We assumed that our days of assembling each issue ourselves were over and we could just hand things off to ChatGPT. No longer would we need to track the industry news, deliberate on which topics to cover, think through the salient points, and then write and edit each issue. Instead, we would open up each issue, just like our readers, and enjoy learning about what has been going on over a cup of coffee. We would use the newfound time to install a firm pickleball court and work on our game. Here in the still shiny opening days of 2023, such a thing seemed entirely possible. Artificial intelligence (AI) experts have been going gaga over the new application. Google, itself a leader in AI technology, is refocusing teams across the tech giant on the threat posed by ChatGPT and related technologies. Educators are likewise in panic mode as ChatGPT appears capable of spitting out solid graduate-level papers. The prolific chatbot has even written a published academic paper and has a second one under review. Microsoft is taking a $10 billion stake in OpenAI (ChatGPT’s creator)  and is reportedly working with the company to incorporate ChatGPT into both its Office product suite (ChatGPT-powered Clippy?) and its Bing search engine (yes, it still exists!). Surely, if ChatGPT can write graduate-level papers, scientific reports, assist with various and sundry emails and corporate memos, and power Internet-scale commercial search engine, it can write a newsletter?
 
We sat down after the holiday break and fired up ChatGPT, ready to hand things over to the AI. We could almost hear the satisfying sound of a pickleball volley. 
 
It did not go as planned.
 

 
At least it got the format right! ChatGPT does some impressive things, to be sure, but the size of the training set is critical. If you feed it thousands of student papers, it can write a passable version of a student paper. If you feed it millions of customer chat logs it can handle routine customer questions. If you feed it millions of scientific and scholarly articles it can write a research paper, although of course it cannot produce the research itself. Sadly (or perhaps fortunately for the sake of our job security), it does not have the training set to produce an issue of The Brief. According to the OpenAI website, it also has only limited knowledge of the past year, making it a poor tool for analyzing current events (this is, of course, not a fundamental limitation in the technology but only of its training sets and processing power). 
 
These limitations aside, ChatGPT is a remarkable technology and one can easily imagine it progressing in sophistication, breadth and timeliness of coverage. Other organizations will no doubt build on the technology and customize it to particular disciplines and purposes. For example, it might be combined with scientific and scholarly editing and author software to help editors or to provide authors with assistance in writing. It could churn out literature reviews, bibliographies, abstracts, and lay summaries. It could further automate translation and language editing software for scientific papers, and help search, curate, and synthesize research, perhaps even helping with meta-analysis. The possibilities seem endless.
 
In ACS Energy Letters, a monthly publication of the American Chemical Society, researchers Gianluca Grimaldi and Bruno Ehrler imagine AI technology acting as as a research assistant, writing:

…we can imagine AI systems proposing new experiments and new descriptions of observed phenomena and arranging data in figures to support their conclusions. An AI system capable of producing original scientific work could revolutionize the whole scientific endeavor─for example, by being less tied than humans to the boundaries of scientific disciplines, bringing multidisciplinary science to new heights.

However, the technology also poses clear dangers. As Grimaldi and Ehrler note, AI technologies are already making paper mills even more prolific and make it harder to detect fraudsters. Authors and editors who rely on emergent technologies will need to guard against errors and mischaracterizations introduced by the AI. As papers become easier to write, authors may be tempted to “salami slice” their research further—producing even more papers from a data set to bulk up their CVs and further straining peer review networks. 
 
Fundamental legal questions also must be addressed before the widespread adoption of AI technology in authoring tools. As the organizers of the prestigious International Conference on Machine Learning (people that presumably understand the technology involved) recently observed in banning papers generated by ChatGPT and other large-scale language models:

As we have already seen during the past few weeks alone, there is, for instance, a question on whether text as well as images generated by large-scale generative models are considered novel or mere derivatives of existing work. There is also a question on the ownership of text snippets, images or any media sampled from these generative models: which one of these owns it, a user of the generative model, a developer who trained the model, or content creators who produced training examples?

(In case you were worried, the lawyers are already hard at work figuring this out.)
 
AI technology also raises questions regarding authorship. Grimaldi and Ehrler ask:

Is an AI-generated scientific text original, despite being the product of training over human-made original work? Who is the intellectual owner of the content? …If an AI system is able to gather its information from the entire literature, will it be able to trace its conclusion back to the most relevant scientific work and acknowledge it? 

These questions become even more complex in the humanities. In the sciences, an original research article is typically a description of data. Clear writing is important, but it is the research and associated data that is most valued. In the humanities, the writing itself (the argument and the style) is usually paramount. 
 
On balance, we think the potential benefits from ChatGPT and its successors in scientific and scholarly publishing outweigh the concerns. Not that the concerns should be dismissed—they are substantive and will need to be thoroughly addressed, as will the legal issues. But the potential improvements in productivity (to say nothing of pure entertainment value and related pursuits) have profound implications if realized.

Springer Nature Reboots Acquisitions

2

Two months ago we wrote about Springer Nature’s first annual report and noted that it acquired few new businesses in 2021, “presumably because it needed to deploy cash to pay off debt, which back in 2018 was €3 billion according to the Learned Publishing article.” The executive team and shareholders are apparently now feeling more bullish, with two acquisitions announced at the end of 2022: Research Square and Cureus.

Springer Nature and Research Square have been collaborating since 2008, with Springer Nature investing in American Journal Experts (AJE) in 2016 and two years later becoming a minority investor in the Research Square platform. In 2020, Springer Nature became the majority owner of both AJE and Research Square, so a full acquisition was a likely outcome sooner rather than later.

The Research Square preprint platform hosts in excess of 196,000 preprints, according to a regularly updated resource on Europe PMC (PubMed Central). It is growing faster than bioRxiv, which hosts 180,000 preprints (as of December, 2022). The Research Square platform’s rapid growth is due to close integration with over 700 Springer Nature journals. “In Review”, which provides authors with a dashboard showing the status of their manuscript in the peer review process, is a useful feature for authors who submit to Springer Nature’s journals. (It’s a bit like the Domino’s pizza app, including the option to order extra toppings in the form of digital language editing—available “in under 10 minutes!”—and data-reporting badges.) Furthermore, authors can showcase that their paper is “under review” at a journal, or in the case of the Nature journals within the Nature Portfolio. Research Square is not a “preprint” platform in the traditional sense. The papers it hosts are less initial drafts of a paper that will be revised and submitted to a journal (such as on arXiv and bioRxiv) and more papers that have already been submitted to a journal and are under review.

The acquisition of Cureus, a rapid-publication open access medical journal that claims to publish papers within 34 days of submission, dovetails well with Research Square’s AJE language-editing services. The Cureus publishing model is unusual: authors can publish a paper with Creative Commons Attribution (CC BY) for free as long as it has “few or no errors” and passes peer review. Other papers either require a light edit ($250–$500 depending on article length, and the number of authors and references) or are currently referred to a “third-party editing service”. 

Cureus has published 32,000 articles to date but has a checkered history, with the dubious honor of having its own page on Retraction Watch. Acquisition by Springer Nature should help Cureus to grow and increase its prestige. Or, as John R. Adler, one of the founders, put it: “What once seemed elusive due to our outsider status is now within reach: indexing in Medline, receiving an impact factor. Everything is now on the table.”

Springer Nature’s two acquisitions reinforce two trends in scholarly publishing. First, the ability to publish quickly and at high volume is commercially and strategically important, especially at the bottom of the transfer pyramid. Second, publishers are looking to monetize as many aspects of the publishing process as possible. Under a traditional institutional subscription model, publication of the certified version of record underpins the primary revenue stream. Now that authors are routinely paying for publishing services themselves, publishers are looking to charge individuals at multiple stages of the publishing process and are building product portfolios, via acquisitions, that allow them to upsell publishing services.

NASA Goes First

3

In December, 2022 NASA’s Scientific Information Policy for the Science Mission Directorate was released and went into effect. This is the first post-Nelson Memo policy on public access to federally funded research, and while compliant with the OSTP’s new directives, is said to have been in the works before the Memo was issued. The NASA policy gives an idea of what is to be expected from US federal agencies as they move to comply with the directives. 
 
For publications, NASA’s new guidance states that funded researchers must make at least the author’s accepted version of published papers available immediately upon publication through deposit in NASA’s Scientific and Technical Information repository. Researchers are encouraged to publish open access (OA), and funding for doing so can be included in research budgets. The policy goes beyond the Nelson Memo in also requiring deposit of technical reports, conference materials, dissertations, and books from funded authors. No licensing terms are specified. Preprints and peer review reports are specifically excluded from the requirements.
 
Open-data requirements are, unsurprisingly, more complex. Data underlying a paper must be released upon publication of that paper, and any further “scientifically useful” data collected during the course of experiments must be released at the end of the funding award period. (“Scientifically useful” is defined as “could be used in either understanding or reproducing a publication or could be used to generate new understanding in the future.”) Data must follow FAIR (findability, accessibility, interoperability, and reusability) principles and be machine readable (structured to allow automated processing). Data must be in non-proprietary, modifiable, and open formats and must include “robust, standards-compliant metadata that clearly and explicitly describe the data.” It is expected that data will be identifiable via persistent identifiers (PIDs) and released under a CC0 (Public Domain) license. Acceptable repositories must align with US Government guidelines and be designated so by NASA. Physical objects, such as laboratory specimens, are excluded from requirements.
 
The NASA policy goes beyond the Nelson Memo in including public access to any software developed during funded research, which follows similar rules to research data, and must be released “under a permissive license that has broad acceptance in the community.” There are no requirements for maintenance or documentation of released software. Additionally, any conference attendance by a funded researcher requires public release of conference proceedings, slide decks, poster presentations, and any other publications produced for the event. Metrics to measure compliance with the policy will be determined at a later date.
 
The requirements around publications are largely what was expected—business model agnostic, but likely to drive authors and publishers toward Gold OA practices. While the policy states that the agency “shall provide funding to comply with this policy”, it’s unclear whether there is any additional funding available or if this will simply come out of existing research budget awards. The data policies show the complexity of such an undertaking and provide routes for excluding work from the policy through administrative approval (one such specified route is, notably, any intellectual property falling under the Bayh–Dole Act). The inclusion of books may give pause to publishers. And the addition of conference materials will create administrative headaches for research societies and other meeting organizers, who will likely need to manage additional complexity regarding conference abstracts and proceedings. It will be interesting to see whether other agencies follow these expanded requirements.
 
The new policy went into effect on December 2, 2022 for all new awards with strong encouragement that researchers with previously existing grants consider complying as well. For those still debating the language of the Nelson Memo, the NASA policy specifically defines the difference between “shall” and “should” and what’s required compared with what is encouraged. Because the other agencies have until 2026 to implement their policies, NASA’s work here will be informative in helping to spot any unintended consequences or complications that arise along the way.

Dealmaking

4

CJK Group, the parent company of Sheridan and Knowledgeworks Global Ltd (KGL), acquires Allen Press. Allen Press was among the last of the independent printers in the US supporting scientific and scholarly publications. 

ASM announced the acquisition of the journal Molecular and Cellular Biology by T&F. C&E had the privilege of representing ASM in brokering this notable deal.

Application developer and provider of infrastructure software Progress has announced plans to acquire MarkLogic.

McGraw Hill acquires medical education video platform Boards & Beyond, further extending McGraw Hill’s footprint in medical education and board preparation. 

Texas Library Coalition for United Action (TLCUA), a consortium of 44 public and private institutional libraries in Texas, reached a landmark agreement for access to Elsevier-published journals. While the deal offers authors at those institutions a 10% discount on Gold OA fees and a 15% discount on hybrid OA fees, this is not a Transformative Agreement, and has been criticized as a retrograde reinforcement of the ‘Big Deals’ of the past. What’s particularly interesting in this deal is a pledge by both parties to develop a pilot to explore the impact of reversion of copyright to authors after some, currently undetermined, time period.

People

5

Markus Dohle is stepping down as CEO of Penguin Random House and will be succeeded by Nihar Malaviya.
 
The World Health Organization has named Jeremy Farrar, Director of the Wellcome Trust, as its Chief Scientist.
 
The Lancet’s Richard Horton was named an Officer of the Order of the British Empire (OBE) in the UK’s New Year Honours List 2023.
 
The American Chemical Society has a new CEO, Albert G. Horvath.
 
The Board of Directors of the American Physical Society (APS) has appointed Randall Kamien as its Editor in Chief, overseeing the Society’s publishing activities.
 
J. Carl Maxwell has been named Vice President of Public Policy for the Association of American Publishers.
 
Alison Mitchell has joined Digital Science in the newly established role of Chief of Staff.
 
Michelle Urberg is the new North American Editor for Learned Publishing.
 
Jason Williams has been named Chair of the Dryad Board of Directors.
 
DeGruyter has named Michael Zeoli as Director of its Publisher Partner Program.

*** 

Former University of Kentucky Press Director Kenneth Cherry passed away on November 18, 2022 
 
Diane Cogan, Vice President of Ringgold Sales, passed away on December 14, 2022.

Briefly Noted

6

The arrest of two Russian individuals for criminal copyright infringement (along with money laundering and wire fraud charges), and the seizing and shutdown of the domains associated with the book piracy site Z-Library, appears to be having ripple effects on other sources of piracy. Vice reports that LibGen is losing mirror sites and moving to the dark web. 

In sad news for readers looking for new books and authors looking for reviews to attract those readers, Bookforum magazine has announced that the current (Dec/Jan/Feb) issue will be its last. 

The Ingram Content Group has begun rolling out Ingram iD, its advertising platform for book publishers.

Capitalizing on the “BookTok” phenomenon, TikTok has begun selling books directly to users.

UK Research and Innovation (UKRI) has issued an OA policy update, with a reminder that any monographs, book chapters, or edited collections that list UKRI funding will need to be published OA as of 2024.

Microsoft and GitHub are accused in a class-action lawsuit of violating licensing terms for open source software in their development of GitHub Copilot, a tool that uses AI to generate software code. The software licenses, like the CC BY license favored for OA publishing, requires attribution of the work’s original creator, a monumental task for a tool that was trained using billions of lines of code. Roy Kaufman suggests this raises some thorny questions for the use of CC BY research articles for text- and data-mining projects.

Several publishers have announced plans to move their entire journal portfolios to OA by applying for Transformative Journal status, including the IEEE and the Royal Society of Chemistry. As noted in the June 2021 issue of The Brief, the Transformative Journal route offers additional compliance with Plan S at little risk, as the status is ultimately non-binding and offers no penalties for journals that fail to eventually “transform.”

Have scientific breakthroughs declined? A new analysis suggests that this is the case. Others have concluded that the results are more about citation behaviors changing over time than any major shifts in discovery.

Liverpool University Press, from the hometown of the Fab Four, launches the first academic journal dedicated to the 20th Century’s most famous band.

***
I believe that at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted. —Alan Turing