Research Resource Commons workshop
PDF version available here.
Workshop on Research and Resource Commons in Scientific Research: Final Report
- American University, Washington College of Law
- November 17-18, 2011
- Michael Carroll, Professor of Law and Director, Program on Information Justice and Intellectual Property, Washington College of Law at American University, Washington, D.C.
In November of 2011, the Washington College of Law at American University convened and hosted a two-day workshop in collaboration with the Creative Commons to develop a strategy for promoting a commons or scientific research and related resources. The workshop brought together interested stakeholders from across the scientific research enterprise: scientists, administrators, librarians, publishers, societies, technologists, lawyers, policy makers, students, funders, and Open Science advocates, including both U.S. and international representatives. This diverse group discussed the current state of policy and technology as it relates to a scientific research commons, and identified key opportunities and challenges, as well as next steps, for the scientific community in general and Creative Commons in particular. These opportunities will inform the next phase of the Science program at Creative Commons and include legal and policy issues, education and technology efforts, and partnerships that will better leverage our efforts going forward.
Rapid advances in information technology, and their uses in the inputs and outputs of scientific research, are far ahead of the legal and policy framework that supports science. A range of initiatives have emerged in the past decade to catch up, including changes in the grants policy at the National Institutes of Health to require public access through PubMed Central to peer-reviewed journal articles arising from NIH-supported research, university-based initiatives to improve open access to the scientific literature and to use institutional repositories as sites for data sharing, and the recent National Science Foundation requirements concerning grantees' data management plans. The time is ripe to review these and other initiatives to assess what lessons can be generalized. In particular, the rapid growth of digital scientific data, the complex status of these data under intellectual property law, and requirements that these data be managed responsibly, suggest that an open, commons-based approach could be particularly useful for addressing these phenomena.
A research or resource commons requires agreement among providers and participants about its legal structure, the technical requirements for its resources, and a shared understanding about how to sustain the commons. Legal issues, usually involving intellectual property or contract law, often arise as researchers, or research funders, seek to build commons or commons-based tools, such as Creative Commons licenses. The objectives for the workshop were:
- To review lessons learned from those who have worked to build or to promote the use of commons structures to support scientific research from within the federal government and from the private sector, including the non-profit sector. This would include review of case studies from existing initiatives to provide open access to the scientific and scholarly literature, attempts to streamline and standardize the sharing of biological materials, and successful data- sharing projects, such as Sage Bionetworks and existing and proposed methods for sharing earth observation data.
- To identify the legal, technical, and cultural requirements for a successful commons, with a particular focus on scientific data. The key themes will be the respective roles of standardization and interoperability at the legal and technical levels necessary for resources to be shared in a commons, whether those resources are literature, data, physical inputs, or others.
- To discuss how the federal government, the university and non-profit sector, and industry can best work together to support existing successful resource commons in science and to create new commons or commons-based tools to improve the speed and efficiency of publicly funded scientific research. Attention will be given to how existing commons standards, such as legal and technical tools supplied by Creative Commons, are currently being used in the sciences and how these might be made more useful with respect to emergent forms of scientific communication.
A commons, in this context, is a standard set of rules by which people access the shared resources, including the infrastructure (standards, protocols, security methods, etc.) as well as the policies and terms of its use (e.g. methods of covering its costs). For the progress of science, we also promote commons that allows for maximal reusability and re-purposability of resources –i.e., the ability to combine large amounts of texts or data for new research purposes often unforeseen by the resource producers but having great potential benefit to science and the public.
The scientific research resources under consideration are primarily the research articles (analysis and conclusions drawn from research) and primary research data produced during the conduct of a research project. Also in scope for consideration are emerging forms of scholarly communication: websites, wikis, blogs, pre-prints, technical reports and white papers, databases, data visualizations, etc. All of these resources may be used in their entirety or in parts, e.g., the reference section of a research article, or a subset of a dataset, so different policies and infrastructure may pertain. Rearch outputs are normally the product of one or more researchers working for a grantee (university, research institution, government agency, etc.) often across international boundaries, so we are interested in a wide range of research materials, research stakeholders, and of international scope.
Historically, formal research publications are copyrighted to the publisher (society, non-profit or commercial), the author and their institution retain no rights, and funders don’t enforce what rights they could retain. This is changing in two ways. First, Open Access publications use different funding models and allow for free public access to articles, but often the publisher retains further rights (e.g. to “mine” the texts or build added-value commercial services from them). Second, a lot of scientific communication now occurs outside the formal research publications, on websites, wikis, blogs, and via informal publications. Many of these are unlicensed and the copyright owner is unclear, others are published under open licenses like CC-BY.
As background for the workshop, a survey of participants conducted before the event helped identify consensus around key questions:
1. Taking account of the ways in which scientific data vary in type and scale, what policies and practices would you say would constitute successful data sharing in science?
- We need to clearly define what is meant by “sharing” (i.e. of what, in what manner and to whom) but in general this will require community buy-in as well as mandates (and enforcement of them), new financial and human resources, and recognition of effort (i.e. academic credit via citation, etc.). Successful sharing will enable reuse of data, implying that it is citable, documented and well-structured.
2. From your perspective, what are one or two of the biggest social, legal, or technical barriers to openly sharing scientific data?
- The greatest perceived barriers were lack of incentives and demonstrations of the value and impact of open data sharing; concerns about data privacy, confidentiality or loss of personal use (i.e., competitive advantage); problems of data heterogeneity, quality and incompatible formats; and uncertainty about legal rights to reuse or re-share data; and the general lack of expertise or best practices to emulate.
3. In which disciplines can we find the biggest successes and failures to share scientific data (e.g. biology, economics, energy, astronomy) and what characteristics of the policy framework, institutional settings or disciplinary norms were most important to success or failure?
- In general, scientific disciplines that have adopted a culture of sharing data are basic (as opposed to applied), have lower commercial potential, do not involve human subjects, and are not classified. Specific areas of success have been in genomics and other -omics, based on ethical and economic principles as well as strong incentives from funders, the geospatial and social sciences, and “big” science such as astronomy and high energy physics where there are large, shared infrastructure costs. Failures can be found in engineering and computer science, ecology and biodiversity, and chemistry (an example of an older discipline in which data was enclosed long ago).
4. What incentive structures most motivate scientists and/or their institutions to share, e.g., academic credit or easier access to data?
- Recognition and credit by institutions and funders was uniformly seen as the key motivator for researchers. Additionally funder and publisher requirements or mandates are good motivators, as well as scientific community values and norms, including ethics and principles such as reproducibility. Finally, demonstrated value of and public demand for data sharing are good incentives.
5. Which resources for scientific research (e.g. research articles, research data, patents, materials, biobanks) are best managed as common resources, and how would you delineate who should have rights of access and reuse to these common resources?
- Ideal candidates for a research commons are datasets that can be aggregated centrally to create a valuable resource (e.g. GenBank). Each discipline has to look at the costs versus benefits of sharing and it is not necessary to share everything that’s possible, but knowing the specific benefits is difficult and sometimes unexpected benefits appear and in unpredictable time frames. Additional pieces of the answer relate to where is it on the spectrum from “raw” or “processed and analyzed” (e.g., the more processed it is the more useful to others) and who its audience is beyond researchers, e.g. students or the public. Some resources with low research value may have high value as educational training data or to support public policy. Linking data to social benefit would improve public awareness of its value, and increase its likelihood of being shared.
- In discussing the survey responses, it was clear that much work is still needed to document the cost/benefit ratio across different types of data in different disciplines. The case for open access to research data is already made in some areas (e.g. genomics) but still very difficult in others.
- Administrators of research institutions are worried about excessive costs, taxpayers want the benefits of research and access to its results but it is uncertain if they see this as justifying additional funding. Solutions to providing access to research resources must be efficient and affordable. Implementing stronger policy at funding agencies will require strong will and credible incentives for researchers and their institutions.
- The workshop was organized are three themes: policy, technology infrastructure, and organizations. This report is similarly organized, with conclusions drawn from each area.
Section I: Policy
The policy goals of government research funders include the advancement of knowledge with high- quality scientific research (i.e. responsible and reproducible), and maximizing the impact of financial investments across the research enterprise, including researchers, grantees, and the public. Workshop participants started from the position that Open Science, including the existence of an openly accessible commons for research resources, is valued by many research stakeholders and should be supported by good policy and range of infrastructure and practices that support its growth and use over time.
Workshop participants reviewed and discussed current policies and activities across a range of U.S. federal agencies, international agencies and specific countries. This gave them a shared sense of current positions on providing access to research in the U.S. and elsewhere, and the types of activities that are influencing both policy and behavior across a range of stakeholders in the research enterprise. Providing a complete picture of current policies, attitudes and activities across such a complex domain is not possible in workshop format so the goal was instead to provide a “snapshot” of a moment in time. The group considered three white papers (provided as appendices to this report), presentations from two major U.S. science funding agencies (NSF and NIH) and various international efforts (CODATA, and initiatives in the EU and China), and had general discussions around new developments and activities in support of improved sharing of research results. These are summarized briefly in this report, along with initial findings and recommendations.
A white paper on U.S. federal policy with regard to research data, focused particularly on unclassified research arising from federal funding agencies (e.g. contracts, cooperative agreements, etc.) was presented and discussed. It examined both government-wide principles and detailed, specific policies for each agency or program that was discoverable by the public Web. The high-level conclusions were that:
- Congress has delegated data-related policy to the funding agencies (although the OMB has a few high-level rules like ‘rights in data’ including copyright, patent, and other rights;
- It appears that in cases where no rights accrue for research data (i.e. for factual data) there is no way for funding agencies to prevent grantees or others from adding contractual access restrictions to the data.
- The NIH has robust and aggressive rules (e.g. for all grants above $500,000) and the NSF is requiring a data management plan that includes policy disclosures;
- In general, individual programs vary from no policy to voluntary sharing to mandates and there is no consistency across agencies. Programs are serving as labs for policy experimentation within different disciplines.
Participants discussed the merits and challenges of this approach to research data policy, whether it was appropriate and at the right level. Many participants felt that data policy cannot be ‘one-size-fits-all’ in science, to allow for an evolving understanding of competing needs across the stakeholders. However it may be possible to create a ‘default’ policy that can be overridden in specific cases, to avoid the common situation of no policy decision due to conflicting interests. But any discussion of research access policies must address incentives for sharing, e.g., how data can be cited for credit in the scholarly record, and how metadata associated with data (citation or other types) will also be shared.
It was also clear that the challenges with data can be more complex than other types of research resources since there are often privacy, confidentiality or security concerns, and the balance of access and other concerns is delicate. Furthermore, both access and privacy requirements for data have deep technical implications for what researchers can actually do with the data once accessed. Policies that require data to be shared must also address how it is shared to achieve these goals.
Many federal funding agencies leave decisions about releasing data to the individual researchers and grantees, but there are a growing number of agencies that are beginning to set policy for how and when data should be shared, in what formats and under what terms and conditions. Across these agencies there exists a range of practice, from strict policy and mandates for open sharing of results within a defined time period to complete silence on the matter. This variability and inconsistency is itself a challenge, since researchers and grantees need to examine the policies and guidelines for each agency (and sometimes each program within an agency) to discover what it requires and how to comply. Universities are well-placed to help research agencies know what each other are doing, and to share the mechanisms and lessons learned from different approaches.
There was consensus that in the context of policy, we need clear definitions of “data” and levels of data (e.g., raw, processed, analyzed, or published). A shared understanding – perhaps a standard taxonomy – would facilitate discussions. Similarly we need to develop a shared understanding of “sharing” and “access” if we intend that a research resources commons will enable reuse of those resources in new contexts.
Most policy related to research resources, particularly data, is made at the private level: private contracts; community norms; licenses and waivers; or explicit agreements, e.g. with publishers. Public policies such as government funding agency mandates are difficult to set and the higher the level, across the spectrum from local to international, the harder the decisions are to make. Given this, when considering at what level a policy should be made it is worth considering how easy it would be to block it, and how close to the local level it could be made without leaving it to individual decision.
Some research data is particularly susceptible to forming a commons via policy, for example, clinical trial data that the U.S. FDA wants available but that data producers may resist sharing. A better course than starting with contentious areas is to choose a research subject that is already a commons, like Antarctica (for which there is already an international treaty to share data). Such a public commons would demonstrate both the value and the costs of aggregating and sharing data in the public domain.
As for software tools, the best investment will be in those that inspire researchers to share rather simply supporting policy mandates. Creating demonstration projects in a research area that includes multiple publicly accessible datasets connected to a specific location seems promising, and could mix both professional and citizen science aspects. Such a demonstrator could then be generalized to other kinds of non-geographic “maps” (e.g., a genome or brain structure). Again, Antarctica seems to be a logical candidate for such a demonstrator.