Is collaboration key to a more connected research ecosystem?

A Research Information webinar explored integrated platforms, interoperability and a new approach to collaboration

WATCH THE ON-DEMAND WEBINAR HERE

There is no shortage of technology promising to transform scholarly publishing. AI-powered research integrity tools, automated editorial workflows and specialist platforms are arriving at an extraordinary pace.

Making them work together is another matter. For publishers, adopting new technology can involve substantial integration costs, complicated migrations and growing dependence on individual providers. Researchers encounter similar frustrations as they move between systems for discovering literature, analysing evidence, writing manuscripts and submitting their work.

Does the industry need more comprehensive platforms? Or would publishers and researchers be better served by an open ecosystem in which different technologies communicate freely?

These questions formed the basis of a recent Research Information webinar, The Future of Publishing Platforms: Competition, Collaboration or Interoperability?, supported by Elsevier and Digital Science.

The panel comprised Martin Rosén-Lidholm, VP of Product and Engineering at ChronosHub; Mark Hahnel, VP of Open Research at Digital Science; Helena Cousijn, Director of Programs and Services at Crossref; and Colin Millerchip, Product Manager at Elsevier.

Connecting the research ecosystem

Helena Cousijn began with Crossref’s vision of the research landscape: an interconnected scholarly ecosystem linking researchers, institutions, funders and research outputs through persistent identifiers and metadata.

Crossref works with 26,000 members worldwide, collecting information that connects articles, datasets and other outputs to the people and organisations responsible for them.

The technology needed to make these connections often exists already, but getting organisations to use it consistently is harder. “I think one issue is the global availability of tools, platforms and solutions,” she said. “I don’t think everywhere in the world people have access to the same thing.”

Even where the necessary tools exist, interoperability may not be a priority: “It’s much more about incentives and importance and prioritisation than it is about not having the technical solutions to achieve it.”

Colin Millerchip, who works on Elsevier’s AI research workspace LeapSpace, approached the issue from the researcher’s perspective.

Today’s market contains numerous overlapping products, he explained. Each may perform a particular task effectively, but researchers must connect them into a workable process. Moving between discovering literature, evaluating evidence, conducting research, writing and publishing creates opportunities for information to be lost.

AI complicates matters further: a researcher might save considerable time generating an answer, only to spend much of it checking whether the result is reliable.

Millerchip sees shared infrastructure and trusted content as essential foundations: “So my view going into this discussion is that we should be collaborating on the sort of infrastructure and plumbing aspects. And if we’re competing anywhere, it’s on the experience.”

Martin Rosén-Lidholm identified another obstacle: established publishing systems often combine their data with the interfaces through which users access it. Introducing alternative tools or creating workflows across several platforms becomes unnecessarily difficult.

For Mark Hahnel, publishing papers more efficiently is useful, but accelerating the production and dissemination of trustworthy knowledge is the larger objective.

Innovation and fragmentation

Hahnel sees enormous potential in AI, particularly its ability to automate activities previously requiring substantial human effort.

There are limits to what current systems can reliably accomplish, however. Large language models trained on general internet content cannot necessarily distinguish sound scientific evidence from unsupported opinion: “There is a way to automate trust, but I don’t think we’re there yet with a lot of the grounding of the knowledge that is being published.”

Hahnel expects technology providers to become involved earlier in researchers’ workflows. Digital Science’s portfolio, including Overleaf and its newly acquired Jenny.ai, reflects an interest in supporting content creation rather than concentrating solely on manuscript submission.

Rosén-Lidholm agreed that specialised AI models could improve reliability, although developing them for individual research disciplines remains expensive.

Among ChronosHub’s publishing customers, the immediate demand is less for additional technology than for improvements to existing systems. Publishers want faster, more efficient workflows without replacing infrastructure they have spent years developing.

Cousijn suggested that fragmentation is not necessarily a sign of failure. New markets need experimentation, with different approaches competing before successful technologies become established.

Millerchip drew a comparison with other developing technology markets. Early adopters often accept inconvenience in exchange for new capabilities. As products mature, expectations change. A clever feature is no longer enough if using it means constantly switching applications or transferring information manually.

WATCH THE ON-DEMAND WEBINAR HERE

The integration problem

Rosén-Lidholm explained that much of the difficulty comes down to how systems are built.

Publishing platforms have traditionally been designed around their own interfaces. Accessing underlying data can involve cumbersome exports, while real-time connections to other applications may be limited.

AI assistants expose these shortcomings; a researcher working on a manuscript might want to consult several specialist services without leaving their document. In practice, that can still mean exporting files, uploading material elsewhere and repeatedly switching systems.

Rosén-Lidholm envisages workflows built around users’ needs, with data accessible independently of the applications used to create or display it: “We need to unlock data and build AI native interfaces to that data.”

Cousijn brought the discussion back to metadata: “It’s really mostly about the metadata that’s registered with it.”

A persistent identifier is only as useful as the information associated with it, she explained. Yet metadata routinely disappears as it passes through publishing workflows. Information collected during manuscript submission may not all make it into the XML, while publishers do not necessarily deposit every available element with Crossref.

Crossref’s service providers programme works with organisations integrating its API to improve the quality of those deposits.

There is a commercial difficulty as well, Cousijn explained. The organisations providing metadata are not always those benefiting most from its reuse. Investing resources in improving it may therefore be difficult to justify internally.

Hahnel offered an example of competitors addressing shared problems. An NIH-funded initiative involving generalist research data repositories, including Figshare, seeks common approaches to disseminating research data and code. Participants cooperate on infrastructure while competing through their services.

“And then they call it coopetition. And then whatever you do above that line is where your business model is, right?”

Competition and vendor lock-in

Millerchip drew on his experience in telecommunications, where operators faced a familiar choice: purchase comprehensive systems from a single supplier or assemble individual best-of-breed components.

An integrated platform offered simplicity and a single point of accountability. Specialist products could deliver stronger individual capabilities, but someone had to connect them. He said: “I think with all of these situations, it depends on what your proposition is as an organisation.”

For publishers whose competitive advantage depends on a particular capability, integrated platforms may impose restrictions. Organisations with limited technical resources might reasonably prefer a comprehensive solution.

The danger is becoming dependent on one supplier, Millerchip added: “Integrated platforms will solve real problems, especially for institutions that don’t have the ability to bring together and build workflows across individual components. But they’ve got to be mindful that it’s not a trap.”

Millerchip pointed to existing commercial cooperation. LeapSpace draws on licensed content from more than 1,000 partners and publishers, providing access to a wider body of research information.

Hahnel highlighted Digital Science’s contribution to shared infrastructure, including GRID, the organisation identifier database that helped underpin the development of ROR. Making resources openly available allows others to contribute improvements, while companies concentrate on the services distinguishing their products.

Rosén-Lidholm was sceptical that established platform providers would willingly make it easier for customers to switch: “I don’t think it’s realistic. And I think you shouldn’t wait for it to happen.”

For vendors selling comprehensive technology stacks, reducing the cost of leaving their platforms is hardly an obvious commercial priority. Publishers can find themselves tied to systems through the expense and disruption of migration.

Rosén-Lidholm’s appeal was directed at the whole industry: “So I would actually make an ask and ask for openness from everyone.”

Building a different ecosystem

Cousijn pointed to Crossref’s history. Publishers needed to link references across organisations, but establishing separate arrangements with every other publisher was impractical, she said. They collaborated to create shared infrastructure. She sees opportunities for similar cooperation in research integrity and research evaluation, where individual organisations cannot realistically solve every problem themselves.

Cousijn suggested that the organisation could have moved more quickly beyond its original emphasis on publications to support identifiers such as ORCID and ROR.

Millerchip expects publishing platforms to become more comprehensive while opening access to their information. AI agents will increasingly perform tasks on researchers’ behalf, he said, requiring machine-readable access to research content and services. Technologies such as the Model Context Protocol (MCP) could play an important role, although standards are still developing.

Trust presents another difficulty, said Millerchip, adding that researchers need access to peer-reviewed publications and emerging findings in preprint servers. 

What does the audience want?

An audience poll provided an indication of priorities.

Asked what should matter most for the next generation of scholarly publishing platforms, approximately 43% selected greater interoperability and open standards. Better integrated, end-to-end platforms attracted around 30%, with approximately 26% choosing simpler, more affordable technology. Just 3.3% wanted more innovation and specialist tools. These were the approximate figures reported during the live poll.

Hahnel connected the result to machine-readable research information. The FAIR principles encourage information to be findable, accessible, interoperable and reusable.

Increasingly, machines rather than people will be responsible for discovering and querying that information, he said, adding that publishing platforms must serve both researchers using their interfaces and automated systems accessing the data behind them.

Smaller publishers and AI reliability

An audience question raised the difficulties facing smaller publishers unable to afford technologies available to major commercial organisations.

Rosén-Lidholm suggested a change in purchasing priorities. Rather than concentrating entirely on features, publishers should examine whether they can access their data and introduce other tools without replacing the whole system.

Cousijn pointed out that the term “small publisher” covers an enormous range of organisations., and that commercial publishing technology may simply be unaffordable for some of them. Open-source platforms such as Open Journal Systems provide an alternative, allowing smaller publishers to participate through established integrations.

Another question concerned AI hallucinations, particularly fabricated references. Hahnel argued that citation-checking technology already exists. The opportunity is to identify problems while researchers prepare manuscripts, rather than relying on checks after submission.

Rosén-Lidholm expects specialised language models to improve reliability, but development costs remain significant. Collaborative investment could help address problems individual organisations cannot afford to solve.

Does the author experience matter?

The final audience question challenged an assumption running through much of the discussion. Do authors care about the experience offered by publishing platforms? Or will they tolerate almost any inconvenience to publish in their preferred journal?

Rosén-Lidholm was unequivocal: “Just treat humans as humans. Be nice.”

Hahnel, drawing on his experience as a researcher, acknowledged that journal prestige can outweigh almost everything else. He said that an author offered the opportunity to publish in a prestigious journal may accept considerable inconvenience or expense because of the career benefits. For researchers choosing between less prestigious journals, the quality of the submission experience could make a difference.

The question also highlighted the sometimes conflicting interests involved. Researchers want to complete their work and publish it. Publishers need systems they can afford, maintain and adapt. Technology providers need viable businesses.

The audience poll suggested where many publishers and other stakeholders would like to see progress. Interoperability, integrated workflows and affordability ranked far ahead of further specialist innovation.

The Research Information webinar, The Future of Publishing Platforms: Competition, Collaboration or Interoperability?, was supported by Elsevier and Digital Science and is available to watch on demand.

Report by Tim Gillett