5

Transitioning to Open Science by Design

SUMMARY POINTS

• Despite the significant progress that has been made around the world, the research enterprise remains some distance from completely open science.
• In order to develop effective strategies for achieving open science by design, it is necessary to take stock of what has worked and not worked around the world. Factors such as costs, researcher incentives, policy and legal frameworks, and publishing strategies need to be taken into account.
• The ultimate goals for an ecosystem that supports open science by design are clear: articles immediately available under gold open access, with data available under FAIR (findable, accessible, interoperable, and reusable) principles, and other research products also available. Additional funding, mandates, and community initiatives can be deployed to push towards open science, but careful planning of stakeholder buy-in will be needed to avoid unintended negative consequences and disruptions.
• Still, there are clear short-term actions that can be taken to achieve further progress, and options for longer-term solutions that can be further explored and pursued.

BARRIERS AND LIMITATIONS

In order to realize the vision of open science by design described in Chapter 4, it will be necessary to develop new tools, technologies, and practices. Researchers will need to see the value in adopting them. Training and reward systems will need to be revamped. Discussion in this chapter covers several approaches that have been proposed or are being tried to overcome the most formidable barriers.

Several barriers to open science are discussed in Chapter 2, including those related to the structure of the scholarly communications market as it has evolved over the years, and particularly developments of the past several decades. The vision of open science by design described in Chapter 4 contemplates that all products of the research process will be available immediately at no charge. This vision conflicts with the traditional subscription-based mode of scientific journal

distribution and related aspects of scholarly communications practices. Many traditional publishers are offering open publication options and new open publishers have emerged, with most using a business model based on article processing charges (APCs) that are paid by the author, the author’s institution, or the sponsor.

Fully open publications are immediately accessible to all researchers at no cost and are available to all researchers under a copyright license that permits them to perform text and data mining or other productive reuses of the literature without the need for any negotiations or further permissions. While some subscription publishers have begun to offer researchers some forms of access for text and data mining and other productive reuses, their terms of access usually impose some restrictions on reuse.

Another important aspect of the transition to open science relates to the availability of data, code, and other research products under FAIR principles. In contrast to the market for distributing articles, the markets for distributing digital research products such as data are unevenly developed.

This chapter covers possible options and pathways for realizing open science by design, taking into account the legal and policy frameworks that apply, and the landscape of organizations and initiatives that are working in this area.

LEGAL FRAMEWORK

This section focuses on how the law treats “openness” as it relates to access and use of scientific information. While intellectual property law is the most common legal regulation of open science, the relevant law begins with the free speech guarantees of the First Amendment of the U.S. Constitution, along with international agreements. Free speech includes the right to speak, the right not to speak, and the right to listen. These fundamental liberties are the baseline condition governing open access to scientific information. When applied to scientific research, they guarantee the right to share and have access to research results.

The Constitution also grants Congress the power to depart from this baseline when creating intellectual property laws consistent with the First Amendment. Intellectual property law balances the public’s right to know against the private interests of researchers to restrict the use of their works for limited times. In the United States, this guarantee provides the legal basis for open science when intellectual property law does not apply. Potentially applicable branches of intellectual property law are: (1) copyright, (2) special database rights in the EU, South Korea, and parts of Eastern Europe, (3) contracts and licenses, (4) patents, and (5) trade secrets. Each of them, except patents, applies automatically and attaches exclusive rights to the protected information.

Although there are cases where private companies have made proprietary data available for research and analysis by the broader community, issues related to proprietary research at companies are not central to open science. Further, although inventions based on university research are often patented, patenting need not interfere with reporting results and making data available. Therefore, issues related to proprietary research that results in patented inventions and trade secrets

lie largely outside the scope of this report. Patents and trade secrets are not covered in what follows. Also, legal issues related to the research use of data generated in other contexts (e.g., social media data) and issues related to the utilization of research results by policy makers are not considered here.

Copyright

Copyright law is the most salient form of intellectual property for this report because it applies automatically to most informational outputs of scientific research, including journal articles, less formal research reports, the organization of datasets, and software. In the United States, federal copyright protection has been granted automatically since 1978, and the requirement that publications carry a copyright notice to maintain protection stopped in 1989. Copyright law is founded on certain science-friendly concepts and imposes no restrictions on sharing the basic building blocks of knowledge—facts and ideas. Researchers rely on this freedom to copy in their daily practice. While, for example, patents apply to specific applications of the CRISPR/Cas9 (Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR associated protein 9) process for gene editing (NLM, 2018b), the ideas and facts underlying the process can be freely built upon. Similarly, raw observational and experimental data are considered “facts” for copyright purposes, free to be shared and reused (NRC, 2009).

Copyright applies to original works of authorship. With respect to journal publications, the “author(s)” who own the copyright sometimes differ from listed authors. Scholarly norms about who receives authorship credit vary by discipline and usually are based on some measure of contribution. The 2016 article that reported the first observation of gravitational waves listed around 1,000 authors, and articles reporting on large clinical trials may have hundreds of authors (Abbott et al., 2016). For copyright purposes, authors are those individuals who wrote the text of an article, created figures or charts, or who otherwise contributed original expression in the work.

Copyright grants the author(s) exclusive right to publicly reproduce the work, distribute copies, display, perform, or otherwise communicate it and make adaptations. When a copyrighted work is created within the scope of employment, the employer is treated as the author and owns the copyright. There is some uncertainty about how this so-called “work for hire” rule applies to outputs of research by full-time faculty, but many institutions have adopted IP policies that address this uncertainty, often recognizing researchers as the authors (and therefore copyright owners) of journal articles they write, datasets they produce or assemble, and software they create.

Most countries also provide authors with some level of “moral right” to their works. These rights are personal to the author and cannot be transferred. Outside the United States, authors have rights of attribution, as well as the right to end attribution if they no longer wish to be associated with the work. A strong version of such rights even gives the author the right to retract a work from publication and to enjoin any further publication or duplication.

Under U.S. law, authors can transfer some or all of their copyright if they sign an agreement to this effect. Subscription-based journals usually require authors to transfer all or part of their copyright(s) to the journal, designating a “corresponding author” who signs on the others’ behalf. This allows publishers to restrict access to paying customers and use the threat of a copyright infringement lawsuit to deter republishing or reusing content without a license. Alternatively, the grant of copyright permission (a nonexclusive license) can be executed less formally. In a case where authors never sign a publication agreement, the publisher holds a nonexclusive license and the authors retain copyright.

The rights of copyright holders are constrained by statutory limitations and by exceptions to the owner’s exclusive rights to certain reuses. These limitations and exceptions vary by country, so that the right to use the copyrighted layer of a dataset—for example, by copying the whole set—without permission depends on where the copying occurs. All countries have a list of uses permitted by law, but these lists vary widely, and the identified uses are often specific and narrow. Countries also create their own exceptions to determine whether a use is permitted, such as the fair use doctrine in the United States and Israel or fair dealing in many Commonwealth countries.

Sui Generis Database Rights—Europe and South Korea

In the EU, certain candidate countries in Eastern Europe, and South Korea, research data may also be subject to a special database right. As frustrating as this may be to a globalized research community, this right could apply to a substantial amount of computerized data downloaded in Europe or South Korea, but not elsewhere.

When sui generis database rights were introduced in 1996, some experts warned that expanded copyright protections, new technologies restricting access to digital content, and database protections could enable proprietary claims to factual matter that previously entered the public domain as soon as it was disclosed (Reichman and Uhlir, 2003). Others asserted that this legal right would be a significant barrier to sharing research data were it not subject to a limitation for noncommercial research. Since then, courts have interpreted this database right in a manner that limits its potential impact on researchers. However, the European Commission launched a review of its Database Directive in 2017, and a 2018 report supporting this evaluation found that European database rights added complexity to data-intensive research and created barriers to making databases open (EC, 2018d).

Contracts and Licenses

When one or more intellectual property rights apply to research outputs, the owner of such rights can grant permission for reuse through a license. In legal terms, a grant of permission is a nonexclusive license. An exclusive license is one

in which the rights holder agrees to give up any rights to use the intellectual property, usually in return for some form of compensation. From a legal perspective, terms of use or other “licenses” fall into one of two groups. In the first group, there is an underlying intellectual property right associated with data that would be violated by the user in the absence of the permission granted by the terms. That is an intellectual property license. Violation of such a license could lead to a court order requiring the user to cease any further use. Damages and attorneys’ fees may also be assessed against the breaching user.

In the second group, there is a collection of data that has no underlying intellectual property right associated with it, such as a large collection of sensor data that is organized in an unoriginal manner—say, chronologically. If one were to download these data from a site with “terms of use” associated, those terms are still enforceable as a contractual agreement, but there would be no intellectual property right to infringe. Enforcing any use restrictions in this second group of agreements is much more difficult because the author of the terms has to prove that the use has caused measureable economic damages.

Although there are policy arguments against enforcing the terms of use in this second group—because they impose use restrictions on data that intellectual property law treats as in the public domain—courts in the United States and elsewhere generally have found these terms of use to be enforceable as long as the basic requirements for voluntary agreements have been met. For example, a Maryland district court upheld a terms-of-use agreement even when a third-party user obtained database access merely by clicking a box to accept, but failed to review, the terms of use. Since the practice is legal and enforceable, it should be a topic for community discussion whether it is ethical or appropriate to condition access to data on agreement to a contract that imposes use restrictions on data that is otherwise free of any intellectual property rights.

Types of Licenses

Rights of use can be shared or granted by several types of licenses. The broadest is the Creative Commons Attribution (CC BY) license, which requires only that the user provide attribution as directed by the licensor. This license is used by open access publishers, including PLOS; creators of open educational resources, such as OpenStax College and Rice Connexions; and a range of other creators.

The owner of IP rights can also grant free permission for use through a nonexclusive license, which applies most directly to data without monetary value. If the data are valuable, the owner may grant an exclusive license, in which the rights holder gives up any rights in return for some form of compensation.

In cases where permission has strings attached, mapping how intellectual property law does—and does not—apply to research data may be of use. For those seeking to understand which reuses of another’s data are permitted by law, regrettably, the answers to the above questions are more context dependent than many

would like. This is so for two reasons. First, the source of all intellectual property rights is national law, so that users’ rights vary by country. Second, as is discussed above, some countries have an additional right that applies to certain factual databases.

Copyright protection can also be permanently removed in most parts of the world if the owner of the rights publicly states the intention to relinquish the rights. Creative Commons provides a tool called CC0 (CC Zero) to remove the rights, and even in countries that deny owners this right, CC0 functions as a license for the user (Box 5-1).

RESEARCH FUNDER POLICIES

This section covers the context for realizing open science by design shaped by the policies and requirements of research funders. The ability to obtain grants from funders to support scientific studies and publish in credible peer-reviewed scientific journals is important for scientists to advance their research careers and receive recognition nationally and internationally for their work. Funders award research projects that align with their values and mission, providing resources to scientists for collecting the data in order to offer solutions to important topics that the funders want to have an impact on. Therefore, the advances that funding agencies and publishers have made are essential in understanding how a transition to open science can be undertaken.

Traditionally, research sponsors do not publish the articles or host the data generated by the work that they support. They may seek to impose certain conditions on awards, which the grantee can accept, reject, or try to modify. In recent years, research funders have taken a more active role in ensuring that work that they support is publicly available, with some going further in their support for open science. Although much of the discussion that follows focuses on U.S. funder policies, international policies are also included because they help shape the global environment for open science.

U.S. Federal Government Policies

Over the past several decades, the federal government has adopted a number of policy changes relevant to open publication or open data. Many of these have affected access to information resources of the federal government itself, or to data that the government produces or uses. Examples include legislative changes such as the Data Access Act of 1999, the Data Quality Act of 2001, and executive branch policies such as the Obama administration’s memoranda on Transparency and Open Government (2009) and Open Data Policy (2013) (NRC, 2009; The White House, 2009; Burwell et al., 2013).

The National Institutes of Health (NIH) was a pioneer in supporting openness in relation to outputs from research that it supports. In 2005, NIH adopted a voluntary public access policy for peer-reviewed literature that resulted from its funding. In 2008, under the Consolidated Appropriation Act, NIH began to require all grantees to submit an electronic version of their final peer-reviewed manuscripts upon acceptance for publication to the National Library of Medicine’s PubMed Central. Articles were to be publicly available no later than 12 months after the official publication (NIH, 2008).

The America COMPETES Reauthorization Act of 2010 called on the National Science and Technology Council (NSTC) to set up a working group that would coordinate federal science agency research and policies related to the dissemination and long-term stewardship of the results of unclassified research. The material covered includes digital data and peer-reviewed scholarly publications, supported wholly or in part by funding from U.S. federal science agencies.

The next significant step toward openness was the release of the memorandum Increasing Access to the Results of Federally Funded Scientific Research by the Office of Science and Technology Policy (OSTP, 2013). The “Holdren memo” directs federal agencies with over $100 million in annual conduct of research and development expenditures to develop a plan to support increased public access to the results of research funded by the federal government. This access includes any results published in peer-reviewed scholarly publications that are based on research that directly arises from federal funds, as defined in relevant Office of Management and Budget (OMB) circulars (e.g., A-21 and A-11).

Several months after the Holdren memo was issued, the National Research Council organized two planning meetings for the federal government to receive public comments (NASEM, 2013c). Over the next several years following the release of the memo, the National Institutes of Health, the National Science Foundation, and other relevant agencies developed their own policies to implement the Holdren memo (NIH, 2015; NSF, 2015). The policies set out requirements for data management plans and public access to scholarly publications to be included in grant applications, though data deposit requirements and publication date requirements varied by agencies (CENDI, 2017). In January 2017, the OSTP published a report to Congress on the progress of these agencies on implementation of their public-access polices (Holdren, 2017). A 2017 analysis of how well the agency plans addressed the themes set out in the Holdren memo related to the availability of research data found unevenness among agencies, with some themes such as digitization/legacy data and digital object identifiers (DOIs) not mentioned or addressed in a significant percentage of plans (Kriesberg et al., 2017).

The Fair Access to Science and Technology Research Act of 2017 is the latest version of legislation that would essentially provide a statutory basis for the policies instituted in the Holdren memo. Bipartisan groups of sponsors have introduced versions of this legislation in both houses of the last several Congresses, but it has not yet passed (S.1701). If adopted, it would provide a stable legal basis for federal support of open science.

Other Open Science Mandates of Funders, Publishers, and Universities

A number of public funding entities around the world have instituted open science policies (ROARMAP, 2018). As might be expected, these policies vary in terms of their coverage, whether compliance is encouraged or mandated, whether article processing charges (APCs) or other costs associated with open publications or open data are covered or not, and so forth. For example, the Research Councils UK policy was adopted in 2012, and was designed to be implemented over a period of five years, with regular assessment (RCUK, 2012). The RCUK policy focuses on open publications, allows for compliance through both immediate open publishing and the use of repositories, and provides for coverage of APCs. The European Open Science Cloud, which was announced by the EC in 2017 with a goal of implementation by 2020, focuses on enabling FAIR data and principles that will underlie data accessibility and stewardship on a Europe-wide basis (EC, 2017a).

Several private foundations have implemented even stronger mandates on open access. For example, as of January 1, 2017, the Bill & Melinda Gates Foundation requires that publications supported by its grants be: (1) deposited in a specified repository(s) with proper tagging of metadata, (2) published under the Creative Commons Attribution 4.0 Generic License (CC-BY 4.0) or an equivalent, and (3) available immediately upon their publication with no embargo period (Hansen, 2017). Further, data underlying the published research should be immediately accessible and open with the foundation paying reasonable fees in order to publish on the terms of OA policy (Hansen, 2017). The Wellcome Trust also demands that results of research that it funds be made openly available within 6 months of publication and provides financial support for those researchers to publish under a CC-BY license (Wellcome Trust, 2016). Other international organizations such as CERN and the United Nations Educational, Scientific and Cultural Organization (UNESCO) have also published open science policies and cover expenses for data sharing. Furthermore, funders are moving from resource provider to knowledge institutions, with increased special funding for programs to understand open science practices and tools (Table 5-1). Several websites have been established so that researchers can check the publication and data sharing policies of funders and publishers (SHERPA/Juliet, 2016; FAIRsharing, 2017; ROARMAP, 2018).

While publishers expect authors to make research data available to the journal or readers upon request for validation and as a supplement to publication, many do not mandate that researchers or institutions provide for FAIR access or long-term curation of the data. Some prestigious subscription journals including Science, Nature, and PNAS, have adopted policies allowing preprint sharing from authors. The SHERPA/RoMEO (SHERPA/RoMEO, 2016) website indexed over 2000 publishers, with 46 percent explicitly allowing preprint posting and 72 percent allowing authors to archive postprints. For example, Science allows authors to immediately post the accepted version of their manuscript on their website and to post to larger repositories such as PubMed Central 6 months after publication.

And the journal Nature allows archiving of accepted articles in open repositories 6 months after publication. For journals that do not formally support self-archiving, authors can submit an author addendum (a template is provided by SPARC, 2016) to allow them to retain rights to post a copy of their article in an open repository.

Universities such as Harvard and MIT have adopted rights-retention open access policies in which faculty members agree to grant their universities nonexclusive reuse rights for future published works. With this policy in place prior to publication, faculty work is archived freely without the need to negotiate with publishers. Many subscription publications offer an open access option that requires authors to pay an APC. There is a significant range in the prices charged for APCs, with many open access (OA) journals charging nothing (Crawford, 2016; West et al., 2014). The majority of OA publishers charging moderate or high APCs offer fee waivers upon request for authors with financial difficulty such as those from low-income or low-middle-income countries. Some publishers (e.g., BioMed Central, F1000, PeerJ) have membership programs through which institutions pay part of all of the APCs for affiliated authors, some institutions provide discretionary funds for author APCs, and some funders also cover fees for publishing in OA journals.

TABLE 5-1 Special Funding Opportunities for Open Research, Training, and Advocacy

SOURCE: McKiernan, 2016.

STRATEGIES FOR ACHIEVING OPEN SCIENCE BY DESIGN

Given the current legal and policy context, how should research enterprise stakeholders work together to facilitate and accelerate the transition to open science by design? This section discusses various strategies and options for achieving open science by design, given the motivations and barriers discussed in Chapter 2, the current approaches discussed in Chapter 3, and the vision of open science by design described in Chapter 4. Transitions to open science should enable the research enterprise and those who utilize research results to reap the benefits—increased reliability of knowledge, more rapid advances, broader participation in science—while minimizing any disruptions. The vision or new status quo should be sustainable in the sense that it needs to succeed over time, create value for stakeholders, and be adaptable to changes in the research and scholarly communications environment.

The committee was not tasked with developing a specific, detailed funding plan and timeline for implementing open science. The committee recognizes the significant cost barriers that remain to widespread implementation of open publication, open data, and open code. The discussion below explores the trends that are likely to affect the adoption of open science, and discusses analysis, policy changes, and options that have been proposed by a variety of groups.

In addition to considering which options might best facilitate a transition to open science, it is important to consider which approaches might be less effective or might have undesirable side effects, such as disadvantaging early career researchers or researchers based in developing countries. Avoiding such missteps will likely be just as important as choosing effective actions.

The committee started with the assumption that all the relevant stakeholders will understand and agree that open science by design is the most desirable future state for the global research enterprise. These stakeholders include public and private research sponsors, universities and other research institutions, and scholarly communicators. The committee also believes that a critical mass of stakeholder organizations will be willing to coordinate policies and funding mechanisms to support both open data and open publications. The committee’s discussion has focused on steps that U.S.-based stakeholders might take, while keeping in mind that no single institution or national body can singlehandedly change the system. Working and thinking globally will help to smooth and accelerate progress toward open science.

Paying for Open Science

Open publication is a complicated mélange of traditional subscription journals, green and gold open access journals, hybrid models, archiving services, and others. In discussing open publishing, we must consider not only relevant costs, sources of funds, policies, and appropriate business models for open publishing, but also how to transition from the current mixed closed-open environment to a model that fully supports open publications. The committee prefers a system that

supports author choice for where to publish. Policy and incentive should drive the system toward open science. Table 5-2 provides a basic outline of dissemination systems based on subscriptions, green open publications, and gold open publications.

Researcher incentives are important. Researcher incentives are very important to take into account when considering transitioning to open publications supported by APCs or by other mechanisms from current subscription-based publishing. As discussed in Chapter 3, bibliometric indicators, most notably the Journal Impact Factor (JIF), play an important role in current research evaluation practices, which affects research funding decisions and reward systems for researchers. The importance of publishing in highly prestigious journals varies widely by discipline and has grown significantly over the past few decades. For example, in biomedical research, Ginther et al. (2018) showed that the most significant predictor of NIH funding is the weighted sum of impact factors of journals where principal investigators publish.

The adoption of open science principles and practices holds the promise of changing incentive and reward systems so that this focus on journal prestige may be reduced. Still, some disparities in the prestige of various dissemination venues might be expected to continue, at least for the foreseeable future, with implications for researcher incentives. As long as universities and funders rely heavily on the signals provided by journals with the highest JIFs, which overwhelmingly tend to be subscription-based, those journals will continue to dominate high-quality submissions, and their publishers will continue to have considerable leverage in negotiating access agreements with research libraries.

TABLE 5-2 Costs to the Research Community in Subscription-Based and Open Access Scholarly Communication Systems

SOURCE: Swan, 2016.

The financial incentives of researchers also need to be taken into account. Currently the average cost of publishing in subscription journals is negligible for many researchers. Thus, the willingness of researchers to devote additional resources from their grant funding or other sources is limited (Tenopir et al., 2017).

Taking the above into account, it is clear that the transition to open science will require a concerted effort on the part of all stakeholders to change researcher incentive and reward systems in ways that place higher value on open practices. In particular, reducing and eliminating the power of bibliometrics in evaluation practices is an urgent task. Funding agencies and research institutions could work together to develop broader measures of scientific contribution such as credit for peer review, data and code creation, replicability, and open publication. On the financial side, co-pays for APC fees would encourage cost competition amongst journals.

Maintaining and strengthening quality review. Quality review and certification of research will continue to be important, and needs to be maintained and strengthened in the transition to a world characterized by open publications. Traditional prepublication peer review, typically performed confidentially by volunteers and organized by publishers, is an important component of the current research dissemination system, in that it provides an expert judgment on the quality and importance of an article and serves as a mechanism to select the articles to fill what has traditionally been a limited number of slots. While the limitations of prepublication peer review in performing these functions are significant and well known, it will likely continue to play an important role, at least in the near future (NASEM, 2017b). It is unclear how review systems of the future will adjust to an open science world where dissemination opportunities are not artificially limited, and where other forms of review made possible by technological and cultural changes come to be more valued.

Strategies to achieve open science might include initiatives to develop and deploy new mechanisms for quality review. For example, authors might publish a preprint that undergoes open peer review and certification before the article is included in an appropriate open journal or online collection. The current system of prepublication peer review has challenges, such as lack of transparency, bias, and exclusivity, that open peer review actually has the potential to improve. Although the scientific community has a long tradition of peer review of journal articles, there is no culture for peer review of other digital research objects, such as metadata for experimental datasets. The success of open science will require new mechanisms to extend peer review to all products of scientific research. There are working examples of postpublication and open peer review, such as F1000 and PeerJ that can be learned from.

Resources and shifts in the distribution of costs are important. Dissemination of research under open science principles requires additional resources. For example, curating and storing data and samples are not without cost. Dissemination of research results requires substantial effort and resources in addition to

the organization of quality review. There is considerable debate over the magnitude of these costs, and the level of income that will be necessary for disseminators to provide necessary services and maintain quality (Van Noorden, 2013).

In open science, revenue can come in the form of grant funding, service fees, dues, membership fees, and donations. There is considerable funding for dissemination already in the system, but it tends to support subscriptions with paywall barriers to readers. Research institutions, the federal government, and other research funders already provide significant financial and in-kind support to the existing system of disseminating research results through journals and other publications. For example, one analysis estimated that Carnegie Research 1 university libraries paid an average of $6 million annually in subscription fees to journals in 2009 (Bergstrom et al., 2014). Some portion of subscription fees are covered by the indirect assessments charged by institutions to research contracts and grants. In addition, institutions and research funders sometimes provide direct or in-kind support (e.g., office space, IT infrastructure, staffing) to scientific societies that publish journals. Likewise, the researchers who serve as the volunteer editors of journals and peer reviewers of articles are typically employed by universities and other research institutions, so that their service represents a significant in-kind contribution by these institutions as well as by the volunteers themselves. Finally, some institutions and funders are covering costs associated with publishing in journals that do not rely on subscription fees. New services and capabilities enabled by open science will require additional resources. The costs of other services associated with open science dissemination, particularly those related to data and the associated software code, may not be covered under current business models. These services are likely to grow in importance in the future and include data analysis, processing, visualization, and mashups with other data.

It is important that the transition to open publication results in a system that serves individual researchers and the community at large at least as well as the existing system. Some will question whether the current system might not be the best available, given that most researchers in the developed world have access to most of the articles that they need and that the system delivers millions of peer-reviewed articles per year. Open publication will need to continue to demonstrate its value. Although the desire on the part of institutions to restrain increases in subscription costs often arises in considering how to transition to open publication, they are distinct issues. Commercial and nonprofit publishers operating on a subscription model may well continue to exist as a key component of the research enterprise. The overarching goal of open science by design is an effective and efficient publication system that ensures openness.

Managing transition is important. Transitioning to open science will involve some uncomfortable changes on the part of stakeholders. It is important that transitions are planned and managed effectively. For example, asking a scientific society to shift its publishing business model as if this could be done quickly and easily is unrealistic. As discussed in Chapter 2, many scientific societies generate surpluses through their publishing activities that support society activities. There

is considerable variety among societies and disciplines in the size of their publishing operations and the extent to which their professional ecosystems depends upon publishing income. Some societies would be severely challenged by the imposition of some types of open publication mandates, especially if they did not include transition provisions. At the same time, research institutions are currently experiencing difficulty in absorbing the steady increases in subscription rates of recent years. There is a need to ensure that institutions can continue to function and perform their functions in the system during any transition.

Reducing or eliminating embargo periods. Central to the rationale for open science are the principles of accelerating discovery and making dissemination of results as effective as possible. Embargo periods work against these principles. Even short embargo periods mean that results are available only to paying subscribers, not to the public, nor to researchers outside a specialty field, nor to search engines, nor to companies, artificially inhibiting progress in an era when scientific progress is accelerating and can in principle be made available immediately. The committee assumes that the ultimate goal of open science is that published results are available immediately upon publication without any embargo period. As discussed above, transitions should take account of the sustainability of stakeholders and their activities. Some disciplines such as physics and economics, where sharing preprints is a long-established practice, are essentially operating on a green open access model with no embargo period today. Other disciplines would face challenges in making such a transition.

Mandates

One possible approach to transitioning to open science would be for all funders to simply mandate gold open publications, perhaps adopting a policy similar to the current Gates Foundation policy, with APCs to be covered by funders as a fixed amount or percentage of the grant, and/or through institutional funds. Some percentage of institutional funding used to support journal subscriptions could be reallocated to support open science projects and infrastructure.

However, mandates, applied naively, can have potentially damaging side effects. It would be difficult for larger agencies, especially those that cover the entire spectrum of research (e.g., NSF), to adopt such an approach abruptly. Some journals that have the reputational value required for advancement in a given field may not currently have open science options. Meanwhile, a number of high-prestige journals important to a field, especially those published by nonprofit scientific societies that operate subscription journals on the bare edge of making ends meet, could suffer disruptions.

There are also reasons to believe that such an approach would produce “winners” and “losers,” and would likely involve several unintended negative consequences. For example, researchers in their role as consumers of the scientific literature would win by having better and cheaper access to publications. Open journals themselves and subscription-based journals with open access options

would benefit from additional revenue. Many authors would gain additional readers. Losers would include the “low-demand” authors who do not have research funding or do not have access to funds from their institutions to pay for APCs (McCabe et al., 2013).

Pushing these results to their logical conclusion, a completely open publication model (without any form of subsidies for APCs) could harm early career, less well-funded scientists and those at less prestigious institutions and institutions in low- and middle-income countries. In addition to early career researchers and those at less research-intensive institutions being “priced out” of publication activity, this might result in less research being undertaken and fewer publications. The Pay It Forward Project (see Chapter 3) estimates that moving to an entirely OA model will be costlier than the current system for research-intensive universities. This contrasts with the current situation where subscription fees paid by less research-intensive universities subsidize the publications of the more research-intensive universities: “Considering both the scholar-as-author and scholar-as-reader roles simultaneously, assessing the net value of OA for scholars appears complicated” (McCabe et al., 2013).

In addition, some are questioning whether a long-term scholarly communications model based on APCs is the best or only answer for science. As discussed above, research funders increasingly recognize that communication of the result is integral to the research process. Without communication, funder investment in research is of no value. As a result, funders are open to covering reasonable communication costs as part of their funding responsibility. This coverage does not have to be in the form of covering APCs. It includes posting papers to preprint servers (some with discipline-organized peer review) and disseminating annotated datasets. Major funders (Welcome, Gates, European Commission) are creating their own platforms, suggesting that the journal’s days of exclusive primacy may be numbered.

Stakeholders might examine how current mandates are operating, including the current NIH policy, the 2013 OSTP Memorandum, and the EC Horizon 2020. Future changes might be aimed at producing measurable advances in open science that feature transparent costs and ensures that key infrastructure facilitates openness and community control.

Community-Based Initiatives

There are excellent examples of well-defined communities publishing in a specific set of journals and negotiating a business model tailored to that community to achieve open science. In this model, an agent representing a community negotiates with the major journals that publish community-specific papers to provide a price considered acceptable to cover the costs of publications in return for making them open upon publication.

An example is the Sponsoring Consortium for Open Access Publishing in Particle Physics (SCOAP3), which, after many years of discussions and negotiations in the high-energy physics (HEP) community, arrived at an arrangement for

open publishing (SCOAP3, 2018). In this case, much of the science community working in HEP is connected to CERN, which represents SCOAP3. Agreements were negotiated over time with numerous publishers in this field; as of January 1, 2018, for example, all HEP articles published by the American Physical Society in their relevant journals have been paid for centrally by SCOAP3. Those articles are now available upon publication. In return, APS reduced subscription rates to libraries. This arrangement will be re-examined 2 years after its initiation.

This arrangement has several notable features: (1) a hybrid model is in effect where certain articles are made available upon publication; (2) a central fund, formed out of a complex arrangement of international agency funds, is used to pay for the cost of publications; and (3) participants used a transition period that moved the system toward an open science environment. Assuming it is successful, it may be renewed and extended to additional parts of the physics community. This arrangement does not include data services, but data are available to the broader community of HEP scientists through data sharing agreements of collaborating scientists that are funded by cooperating science agencies.

It is important to note that in this arrangement, authors are not required by their funding agencies to publish in open access journals, nor are any journals required to switch to open access. Authors can still choose where to publish, and journals can decide if they wish to join the agreement and offer similar services. In principle, a HEP author might still choose to publish in a subscription-based journal that is not open, but given these incentives and community norms it is highly unlikely that one would choose to do so.

Issues Raised by Data and Related Services

Open science means more than open publishing. Researchers need to archive data and code and scientific collections associated with preprints and publications. The committee started with the assumption that the desired end state is for all data underlying reported results to be openly available under FAIR principles, with disciplinary standards in place to determine which data should be preserved over what time period and clarity over how the costs will be covered. The economic issues around data are very different from those around publishing, largely because there is already a mature publishing industry with established funding sources and business models in play that evolved over centuries, whereas data services are not well developed nor funded in many fields.

Most data in repositories today are not available under FAIR principles, and the complexities of realizing this will entail significant costs. Making data FAIR is a difficult task for investigators, and substantial public investment is going to be required to change the current situation. Making data “findable” is going to require better standards for metadata; new ontologies for the vast majority of scientific disciplines, which currently lack standardized, granular terms that can be used by data search engines; and new tools to enable investigators and curators to author scientific metadata that are sufficiently comprehensive and standardized so that search engines can locate appropriate datasets with adequate precision and recall. Making

data “interoperable” and “reusable” can only be achieved if the data are annotated with comprehensive, standardized, high-quality metadata. Again, the absence of necessary metadata standards, appropriate ontologies, and easy-to-use annotation tools is a significant barrier. There is a misconception in the scientific community that simply putting experimental datasets in the cloud will make them FAIR. Scientific publications become findable and useable to others only when they are well indexed; scientific data require nothing less.

The practices of data curation and dissemination lag behind article publishing in most fields. For data curation and services to be provided routinely, significantly more agency funding will be required. This is absolutely necessary to support the vision of open science. With over 2.5 million peer-reviewed journal articles published each year with a projected annual growth rate of +3.5 percent (EC, 2012), completely new funding sources, business models, and even businesses will be needed to support not only storage, discovery, access, and delivery of data, but also new solutions that could entail curation, replication tools and services, and the like.

The premise of open science by design is that scholarly articles and associated datasets should be open and available for others without paywall barriers, and that the agencies and universities that support and perform this research should consider the cost and curation of these results to be part of the cost of performing research. However, the services of analysis, manipulation of data, visualization, and so on do not all have to be borne by the funding sources. Such value-added services might be provided by for-profit businesses or by nonprofit organizations. Service providers might compete for subscribers in an open market.

When creating a change, some steps might be eliminated, new steps might be added, and the existing relationship between value, revenue, and cost will change. For example, in the recorded music industry, disruption was caused by changes in recording technology—phonograph to LP to 8-track to CD to streaming and packaging—and purchasing a curated form (an album) gave way to single-track selection and adoption. The “unbundling” and “cord cutting” phenomena seen in cable television represents a similar transition. In the current publishing model, data might be thought of as part of the new “listening” stream of open science, and sharing of open data is perhaps analogous to the “sampling” of music.

Some of the conditions needed to ensure that the necessary technologies and infrastructure exist to realize open and FAIR data and code on a universal or near-universal basis are described in Chapter 3 and Chapter 4. Important elements include: (1) clarification and standardization of data management plan requirements with enforcement mechanisms, (2) training/assistance in data/code archival best practices, (3) development of disciplinary guidelines for the resources that need to be preserved, and (4) resources to support these activities.

It is important to understand the barriers to realizing open and FAIR data across the sciences. At the same time, it is important to recognize and learn from successful efforts to create community data resources described in Chapter 3, such as the National Center for Biotechnology Information (funded publicly by NIH)

and the Sloan Digital Sky Survey (funded privately by the Alfred P. Sloan Foundation). These efforts have accelerated progress in their respective fields.

Changes in the Business Environment

An additional issue to take into consideration in developing strategies to achieve open science is the shift in the business environment around scholarly communication. For example, commercial publishers have undertaken significant horizontal and vertical integration in recent years. As a result of mergers and acquisitions, just five firms now publish over half of the world’s peer-reviewed literature (Lipton, 2006). In addition, commercial publishers are acquiring important pieces of the scholarly communications infrastructure, such as preprint servers and institutional repositories, and expanding data archiving and analytics services associated with their journals (Posada and Chen, 2017; Schonfeld, 2017). For example, Elsevier is working to capitalize on movement toward open access by integrating services and tools for researchers and institutions that address needs from “the research design and grant application stages through laboratory research and to and even beyond publishing” (Schonfeld, 2017).

At the same time, a number of libraries and library consortia around the world have taken a harder line in negotiating with commercial publishers (Schiermeir, 2017). The emergence of national consortia with greater bargaining power than individual universities means that commercial publishers may have less room to raise prices (Normile, 2018). The “big deal” pricing strategies of journal publishers have affected the market for research journals, as described in Chapter 2. The growing trend of “big deal” journal cancellations and the rise of availability of free, legally vetted copies of manuscripts (through use of oaDOI and other tools, such as UnPaywall, the Open Access Button, etc.) signals growing support for open science.

Short-Term Steps

Given these trends, what steps might the research enterprise and its stakeholders take to move closer to open science? Possible steps might be framed in terms of short-term and long-term actions that uphold the principles of open science.

Realize Universal Green Open Publication

While an immediate, universal gold open publication mandate might have negative unintended consequences, a universal green open publication mandate would have a number of beneficial effects. This would not be the end solution, since green open publishers might not have final versions of articles and would include material that ends up not being quality-reviewed at all, but such a mandate would be an important step toward a fully open system. There is some useful experience and precedent. For example, the advent and growth of arXiv and other

preprint servers over the last quarter-century has expanded access without disrupting scholarly communication in physics and astronomy. An appropriate embargo policy would need to be determined, such as the current Science policy that allows immediate self-archiving and deposit in an open access repository after 6 months.

In order to support such a mandate, funders will need to articulate a desired rights retention or licensing designation in their funding terms and conditions. Likewise, institutions would need to include rights-retention provisions in their campus policies. Researchers would be encouraged to use available tools, such as author addenda, to retain rights to research outputs. The University of California’s University Committee on Library and Scholarly Communication issued a Declaration of Rights and Principles to Transform Scholarly Communication, which is one example of a supportive institutional approach (UCOLASC, 2018).

Green open publication and preprints are not synonymous. Other models for green open publishers that incorporate more elaborate quality review exist and can be expanded. In addition, green open publishers often require a support model that includes institutional membership funding and grants from foundations.

The committee believes that a comprehensive plan to bring the market to a world of fully open science could start with an approach along these lines, but it would need to be accompanied by additional steps as discussed below.

Devote More Resources to Data Management and Other Open Infrastructure

Additional resources that will be required for open science could come from several sources. Although a rapid repurposing of the fees that libraries devote to journal subscriptions might be difficult to manage, the idea of devoting some of these funds to support open infrastructure (the 2.5 percent commitment) has gained some currency within the research library community (Lewis et al., 2018). The Open Research Funders Group and its members are another source of funding for open infrastructure (ORFG, 2018).

Some voices have stressed the importance of community control of open infrastructure: “Everything we have gained by opening content and data will be under threat if we allow the enclosure of scholarly infrastructures” (Bilder et al., 2015).

Adopt Approaches to Evaluation and Reward Systems That Avoid Misuse of Bibliometric Indicators and Value Open Science

As discussed above and in Chapter 2, approaches to evaluating research and researchers that misuse bibliometric indicators constitute a significant barrier to more rapid adoption of open science practices. Quite a few funders and journals have signed the San Francisco Declaration on Research Assessment (DORA, 2013). Some funders are taking specific steps such as limiting the number of citations that can be included in a proposal biosketch, and allowing preprints and

other interim research products to be included in applications and reports (NIH, 2017a; NSF, 2016b).

In addition to addressing deficiencies in evaluation practices, research institutions and sponsors could make efforts to visibly reward open practices as a direct effort to improve progress towards their core missions. The idea is that open science is not an “add-on” for these stakeholders, but an important enabling strategy to become more effective.

Strengthen Consortia of Libraries and Other Research Consumers

An important element of controlling costs in a transition to fully open science will be to ensure that stakeholders are not simply adding additional resources to what they are already paying for access to research results. The NorthEast Research Libraries Consortium coordinates and negotiates terms for 30 core academic research libraries.

Greater transparency in the prices that research libraries are paying for subscriptions would also be of wide value to the community (Ploeger, 2017). Although some commercial publishers routinely use nondisclosure agreements, public institutions are often subject to state freedom of information laws, and disclosure of terms can be pursued through this mechanism. The UCOLASC’s Declaration of Rights and Principles to Transform Scholarly Communication, cited above, includes 18 principles regarding rights and principles when negotiating with publishers during journal license renewals (UCOLASC, 2018). If adopted, it will influence future journal license negotiations with publishers.

Possible Long-Term Actions

Depending upon the degree of progress made in taking some of the short-term actions outlined above, a more significant set of steps might be explored by stakeholders. Such an approach might combine more far-reaching mandates and other coordinated policy changes, a defined transition period, and a “burst” of funding to cover costs associated with the transition, including those associated with both transitioning to open publishing (e.g., with a temporary hybrid period) and provisioning additional data services (e.g., minimally, new services for data associated with published articles).

The transition period is likely to require some years for the market to adjust, analogous to the period when the music industry experienced a dramatic realignment of revenue and cost reduction stimulated by new delivery models. It will also require policies coordinated across agencies and countries that will differ for science communities that currently operate in different ways. Likewise, the transition is also likely to need a temporary infusion of funds to initiate and cover the transition. Sources of this “burst” funding could come from philanthropic investment, federal agencies, and universities. Since this funding would be designed to incentivize different behavior, clear benchmarks and requirements would need to

be developed. Government participation is likely to require extensive time due to the processes of approval and budgeting.

Communities that have already begun this process may experience little difficulty, but others may face extreme disruption as the market sorts itself out.

Define and Frame a Commitment to Open Science

A more aggressive pathway toward open science might involve one or more commitments on the part of stakeholders to achieve openness by a given date in the future. Given the time that has passed since the 2013 OSTP memorandum, it might be worthwhile to revisit and update this key federal policy. Also, Congressional passage of the Fair Access to Science and Technology Research Act would constitute a stronger commitment to open science than the current policy based on an executive branch memo. The OA2020 statement, established at the 12th Berlin Open Access conference in 2015, is an international initiative aimed at moving the community toward open science on a global basis (Samberg et al., 2018). Its support is not very broad, however. As of April 2018, 107 scholarly organizations worldwide have officially signed the Expression of Interest, including the University of California Los Angeles and UC Riverside (Open Access 2020, 2018).

Expand Voluntary Market Initiatives

The SCOAP3 initiative described above is an example of a voluntary agreement, negotiated by a central organization, that makes certain articles in numerous journals available without an embargo period. This is a big step toward an open science model during a transition period, but it is limited to a specific sub-field of physics and to specific parts of existing journals; the other parts are still behind subscription paywalls. These ideas illustrate how one might go further and coordinate policy across different countries, different funders, and different fields to incentivize the creation of conditions for an open science environment. The goal is to transition to open science without resorting to simple mandates or complex negotiations around a central organization in every subfield (which may nonetheless still be helpful in transitioning to a fully open science model).

Generalizing from the SCOAP3 example, journals might be willing to switch voluntarily to a business model of gold open access if enough funders that cover a given domain (say, physics, computer science, or civil engineering) all agree to an open science policy and provide funds to support it. However, several features are required for such an approach to work: (1) critical mass—support would likely need to span multiple countries; (2) regional flexibility—different countries have different models for funding research and publications (e.g., direct vs. indirect funding of subscriptions); and (3) discretionary waiving of APC costs—there is a need to account for authors without support whose papers merit publication in important journals with an APC-based open science business models.

It is important to note that the complexities of the SCOAP3 arrangement are not all presented here, and that it benefits from the concentration of the HEP community around one international, centrally administered institution (CERN). Adapting the model to other communities may not be straightforward.

Develop Concepts for Transitional Funding

As discussed above, there is currently significant debate over the desired future state of scholarly communication, with some assuming that the future will be dominated by open access and hybrid journals that charge APCs, and others resistant to a future where APCs are the primary mechanism for financing scholarly communication. Whichever approach is ultimately pursued, a burst of transitional funding might be provided to support researchers and institutions that lack the resources necessary to cover APCs or other costs, to support societies that rely on subscription income from their journals to support general society functions, and to launch a new scholarly communications infrastructure that does not rely on APCs. In addition, availability of funding for publication might be made contingent on publishers meeting certain conditions.

During the transition period, additional funds would likely be needed to initiate the movement of the market toward open science so that costs for new approaches could be paid while subscriptions are still also being paid. One issue encountered in other countries that have mandated open publications is that the average APCs of fully open journals are about half as much as those “hybrid” subscription journals that offer single articles in open form (Swan, 2016). To the extent that APCs are supported, a key to this transition would be to ensure that publishers are only eligible for hybrid APC charges if they lower subscription fees as an increasing fraction of their articles are open. A “hybrid-to-open process” would need to be carefully monitored. There is a need to prevent “double dipping”—accruing APC and subscription revenue from the same articles—on the part of publishers (Swan, 2016).

Such an approach would require negotiation and monitoring, but lessons learned by SCOAP3 may be useful. For example, agencies might agree to pilot a switch to such a voluntary market in some specific disciplines, with the goal of opening the entire market after a period of, for example, 5 years.

Explore New Standards and Governance Approaches

An ambitious transition might require new approaches for managing open science within the community at large. This can be complex or very simple. Federal agencies would need to work with research institutions and other stakeholders, perhaps through a new coordinating mechanism.

The research community might also need to create and accept standards that can lower operating costs, provide the structure and guidance that allow for smoother operations, and speed the process of getting data into an open science

format. Standards developed by the community will speed the growth of markets and numbers of users which will justify and allow for investment by service and support organizations in new systems, services, and products. It is also possible for researchers to support services themselves at a low barrier of access as they do for commercial services. Transition to this model would need to overcome considerable resistance from many in the community. As described in Chapter 3, openness varies significantly by discipline, with the highest levels in biomedical research and mathematics, and lower levels in chemistry and engineering (Piwowar et al., 2018). Common issues in different disciplines are availability of infrastructures, policies and standards, and culture. There is a need for raising awareness within different disciplines of the importance of setting standards to move towards open science.

With the creation of standards, a development road map can be created to allow existing services to be extended and new solutions to be developed. A community-based model could ensure that FAIR processes are followed, requirements from the community are worked on at the appropriate time, and resources are allocated to the needs deemed most urgent. A development road map done clearly, transparently, and with appropriate context will benefit the funders, recipients, and consumers of open science.

Provide for Training

Open science will introduce new steps and new processes. As discussed in Chapter 4, training in open practices has to be an important part of the rollout and of ongoing education of all stakeholders, including researchers, librarians, universities, and funders. Training can also help gain supporters and advocates to improve the system toward an open science enterprise.

Ensure Fair Pricing

The open science workflow will incur costs at various points, some of them at service or vendor prices. Creating an ecosystem in which there is a monopoly or single supplier could cause costs to rise without being checked. Conversely, too much competition will fragment or balkanize the market so no individual or group of stakeholders can achieve a critical mass associated with low-cost behavior. Fair pricing also allows for capital that can be applied to investment in future development (features, enhancements, or even disruptive innovation).

Cover New Service Costs

While traditional funding sources can help support publishing in an open science model, additional costs will be needed for new services, particularly those of data storage and analysis. In the special cases of very large data science projects

(e.g., LSST, SDSS), data management and services are fundamental to the projects themselves, and plans to cover costs are usually in place from the outset. But for the majority of research projects, across virtually all fields, the practice of data stewardship is not supported, and significant funds to cover additional data services, both disciplinary and interdisciplinary, will be needed.

For new areas required to support open science, the three main funding sources—government agencies (e.g., NSF, NIH), private foundations (e.g., Gates), and universities—will need to collaborate on several strategies: covering costs through a mixture of local (university) data stewardship; project-based stewardship, in which the receiver of a grant spends a portion of a grant on the new services; and centrally funded national data services that together support open science.

Strengthen Community Leadership

We can already see that the answer depends on community leadership. It is clear that the technology enablers of open science are causing disruptions in many markets, and that open science is just one of many outcomes. But as science is largely performed and funded by governments, foundations, and universities, this “market” is not able to adjust as quickly and freely as other fields such as the music industry (although its transition might still be traumatic, e.g., publishers may go out of business, early career open science practitioners may have difficulty with promotion and tenure, etc.). Therefore, leadership will be needed to develop, coordinate, and implement policy in order to ensure that the disruptive transition to fully open science is orderly and complete, and the goals of open science can be realized effectively, affordably, and quickly.

In the U.S., a possible example for community leadership might be an organization experienced in setting standards and developing technologies, such as the National Institute of Standards and Technology (NIST). While the committee does not make a formal recommendation with regard to NIST, it suggests that such a combination of skills would be appropriate, as long as the entity is tasked at the level of OSTP with working across agencies and foundations. The task of developing a set of coordinated policies that accelerate and manage the transition to open science publishing is indeed a daunting one, whomever is charged with its execution. In addition, community groups would need to be established to advise and guide the processes, which will vary for communities in very different stages of open science practice. For an undertaking of this scale, international coordination would be required, especially with other OECD nations that have already begun this process.

In addition to these considerations, for a voluntary market to operate effectively there must be mechanisms to keep costs down. If coordinated policies from funders of science result in simply moving funds from covering subscriptions in libraries to APCs—say, from a general fund—competition among journals may not be enough to limit price hikes. Authors would have little incentive to choose

lower cost journals if costs were fully covered by a central fund. One strategy would be to require APCs to be paid partly by a central fund and partly from an author’s grant. This would ensure that authors are mindful of costs and incentivized to choose lower-cost journals. Together such policies could enable open publishing to operate voluntarily. After a transition period, during which adjustments are made to the funding mechanisms (e.g., how much is allocated centrally vs. from a grant, what revenues might be generated by value-added data services, etc.), the market should become more efficient and driven to lower cost.

This page intentionally left blank.