As we look back on the last decade of research and forward to the next one, there’s no question that the world of scholarly publishing is changing. Over the last ten years, academic journals have officially gone digital, with the latest STM Report stating that “virtually all STM journals are now available online, and consequently the vast majority of journal use takes place electronically.” At the same time, the vision for a predominantly open access (OA) publishing landscape has shifted from a possibility to a probability in the opinions of many. A 2017 Springer Nature survey of 200 professional staff working in research institutions around the world found that over 70% of respondents agreed scholarly content should be openly accessible and 91% of librarians agreed that “open access is the future of academic and scientific publishing.”
Changing publishing norms and expectations have given rise to a new wave of publishing technologies and standards that we’ll be riding into 2020 and beyond. Where are the biggest scholarly publishing advances occurring? In this blog post, we look at five trends to watch in the new year.
We’re starting off with two of the biggest buzz words both in and outside of academia right now — machine learning and artificial intelligence (AI). First, a quick point of clarification — while the terms machine learning and AI are often used interchangeably, they are not exactly the same thing. What’s the difference? At the highest-level, machine learning is an application of the concept of AI. AI is the theory that machines can be made capable of performing tasks that require human intelligence. Machine learning is a form of AI based on the premise that computers can be “trained” to interpret and “learn” from data sets and machine-readable text. Machine learning programs underly some of the latest smart technologies, like self-driving cars. They are also fostering some promising publishing innovations.
How could AI and machine learning be applied in academic publishing? Two areas that stand out are content curation and interpretation.
In the realm of content curation, AI and machine learning have the potential to change how publishers and aggregators organize content and deliver it to readers. New machine-learning applications are making it possible to automatically create groupings of related content and even make smart recommendations to readers based on what they’ve read before. Moving to content interpretation, AI and machine-learning presents opportunities for scholars and publishers to more quickly and easily understand and make connections between research. For example, AI could be used to help find relevant peer reviewers for manuscripts by identifying scholars working in similar research areas and recommending them. Tech startup UNSILO is piloting AI tools in all of the aforementioned areas. Another AI startup helping scholars and publishers make connections between research is Scite.ai, which analyzes article citations to see if they are supporting or contradictory. At Scholastica, we’ve also begun experimenting with machine-learning to improve the value and efficiency of our typesetting service.
How significant of a role AI and machine learning will play in the future of scholarly publishing is yet to be determined. As the Scholarly Kitchen article, “Ask The Chefs: AI and Scholarly Communications,” points out, there are many questions to be answered, such as the foundational data needs for AI to work effectively at scale. Jasmine Wallace, Peer Review Manager at the American Society for Microbiology, noted: “we’ll have to provide enhanced metadata to maximize the effectiveness of AI.” Other commenters debated the strengths and limitations of present and potential AI applications.
We’re not through talking about machines just yet. Another aspect of online publishing that’s becoming a focus area for stakeholders across the scholarly communication landscape is interoperable metadata. The transition to online publishing has resulted in a plethora of digital systems for information to pass through — from indexes to archives to aggregators, and the list goes on. To recognize and organize scholarly content, these different systems need to be able to process information in computer markup languages — that means machine-readable metadata. And with so many interrelated systems, there is a growing need for machine-readable content standards to make metadata more accessible, reliable, and reusable.
As NISO Executive Director Todd Carpenter explained in his October letter, “There are other elements of serving content to machines that do matter, sometimes a whole lot. Consistent file structures streamlines content ingest and transfer among parties in the supply chain. Metadata describing the content can impact discovery, sales, or usability of a particular file. The presence of accessibility features might facilitate or inhibit (if they are missing) readability by some devices in certain circumstances.” And (you guessed it) richer, more interoperable metadata could have a direct impact on the speed and accuracy of present and potential AI and machine-learning initiatives.
In his October letter, Carpenter shared that NISO is in discussion with ANSI and ISO about developing more reliable and transferable machine-readable standards. Another notable initiative around improving metadata quality and interoperability that’s underway is Metadata 2020. The goal of Metadata 2020 is to bring together stakeholders in scholarly communication to discuss how metadata is used throughout the research lifecycle and to develop recommendations to improve metadata, including developing shared data mapping standards and expectations.
As noted, there is growing consensus within academia that the majority of scholarly content will be available OA in the future — but how to reach that end is still a matter of debate. The announcement of Plan S in September 2018, an initiative by a consortium of national and international research funders to make research fully and immediately OA, sent shockwaves throughout academia. 2019 saw the release of the revised Plan S guidelines with some significant changes, including an extension of the Plan S deadline to January 2021, a clearer Green OA compliance pathway, and greater flexibility around non-derivative copyright licenses. What remains the same — and has been a matter of significant debate — is that Plan S will not acknowledge hybrid OA as a compliant publishing model.
In response to concerns raised by scholarly societies around the feasibility of transitioning to full and immediate OA publishing without compromising their operational funding, Wellcome and UKRI in partnership with ALPSP launched the “Society Publishers Accelerating Open Access and Plan S“ (SPA-OPS) project to identify viable OA publishing models and transition options for societies. The final SPA-OPS report was released in September of 2019, encompassing over 20 potential OA models and strategies as well as a “transformative agreement toolkit.” Among the most promising OA avenues identified by SPA-OPS include:
- Transformative agreements
- Cooperative infrastructure and funding models
- Author self-archiving
- Article transaction models
With these new resources available and the Plan S deadline getting closer, 2020 is likely to be a year of experimentation around different OA publishing approaches.
Since the San Francisco Declaration on Research Assessment (DORA) recommendations was released in 2012, a growing number of scholarly institutions and funders have been calling for alternatives to the Journal Impact Factor (JIF) driven research assessment culture. As DORA points out, the JIF “was originally created as a tool to help librarians identify journals to purchase, not as a measure of the scientific quality of research in an article.” DORA argues that research should be assessed “on its own merits rather than on the basis of the journal in which the research is published.”
DORA has gained steady support over the years. Now, with new OA publishing initiatives like Plan S calling for changes to the publishing incentive system, to support the development of OA journals and curb corporate control over the majority of high-impact titles, it seems that DORA implementation may start to speed up. DORA recently posted a list of top 10 advances in research assessment that includes examples of organizations that have committed to introducing concrete research assessment reforms in 2020.
Last, but not least, another trend we see for 2020 is increasing use of preprint servers, and not just for posting articles before formal publication. In recent years, we’ve seen a growth in preprint servers and new uses, including archiving final versions of articles to make them Green OA; publishing research datasets, code, and supplemental materials; and even publishing academic journals via preprint overlay models.
With recent OA publishing mandates and open data initiatives, it seems that preprint usage will only increase. As noted, Green OA via author self-archiving is one of the transitional models available for publishers to comply with Plan S. In recent years, there have also been more examples of preprint overlay journals that host their articles on preprint servers like the arXiv to reduce production costs. The Center for Open Science has also been promoting preprint usage for posting and sharing open data through its Transparency and Openness Promotion (TOP) guidelines. To reach the highest level of TOP compliance research data, “must be posted to a trusted repository.”
Only time will tell what the next decade has in store, but one thing for sure is we will likely see greater innovation in all areas of scholarly publishing. That’s what we’re focused on here at Scholastica, to make research more widely accessible and affordable. We’re excited to see how the digital publishing and OA landscape will continue to develop in 2020.
What other scholarly publishing trends are you watching this year? Share your thoughts in the comments section!