3

Establishing Collaboration Principles

To set the stage for the later efforts to talk about possible research data collaborations, an early session was devoted to identifying principles that could govern such collaborations. To do that, summit participants were separated into groups to carry out parallel discussions and then later reconvened to share the results of those discussions with the entire summit.

Each group was asked to answer the following three questions:

1. In your experience, what are the characteristics of successful cross-sector research data collaborations?
2. What are 3–4 guiding principles that would advance cross-sector research data collaboration? Rate them by ease of adoption high/medium/low, with high being the easiest to adopt.
3. What potential pitfalls do we need to mitigate?

To get the discussions started, the breakout groups were also provided with two previously established sets of principles from other groups interested in scientific research.

This chapter describes the set of principles that were identified through this process. The principles discussed during the breakout were reviewed when the breakout groups reported back to the summit plenary, and so individual participants are generally not identified in this summary.

CHARACTERISTICS OF SUCCESSFUL CROSS-RESEARCH DATA COLLABORATIONS

What characterizes a successful cross-research data collaboration? Participants in the breakout sessions came up with several characteristics that in turn informed their discussions about identifying principles for such collaborations.

Trust is essential. While it is important to spell out the various parties’ commitments, responsibilities, and benefits in contracts, memoranda of understanding, or other formal documents, it is never possible to account for every possible contingency. One must be able to trust that the other parties will not only fulfill their obligations but will also behave reasonably and fairly in all situations. Trust makes it possible for all parties to put in their best efforts without worrying that they may be taken advantage of.

Communication goes hand in hand with trust. Clear communication helps build trust, while trust makes communication easier. However, breakout discussion participants pointed out that trust does not solve all communication issues, and there is much work that needs to be done to make sure that communication among all parties is as smooth and effective as possible. One aspect of good communication is establishing a shared understanding of acronyms and vocabulary and a shared definition of data; this is particularly important for collaborations that involve experts from different disciplines or different cultures.

Given that collaborations are likely to include people from different backgrounds, breakout participants saw it as important that these collaborations represent and benefit the diversity of interests, identities, and communities they are intended to serve. All voices need to be represented and heard, and areas of mutual benefit need to be identified.

Collaborations are more effective when they acknowledge and mitigate tensions between sectors. Communication is vital in exploring and understanding cultural differences that might not be apparent at first. Individuals need to understand the culture of a sector they are working with, including its motivators and stressors.

One way to form bonds and strengthen connections among the different groups in a collaboration is through a shared vision and understanding of goals. It is not necessary that the groups share goals (although it can be helpful), but it is important that they understand what each other want and need out of the collaboration and how it will benefit them. The research questions being asked need to be clear to everyone, and the time-

line be mutually agreed upon and understood. With this sort of mutual understanding—and, hopefully, common goals—it is possible to create a community of individuals and groups that are all working toward the same things. This way, a successful project leads to collective benefit.

Beyond these mostly cultural considerations, participants identified several technical considerations that play a role in successful cross-sector research data collaborations. For instance, it is important to establish common standards for metadata and provenance. Without those common standards, it is difficult, if not impossible, to maximize the benefits of a cross-sector collaboration.

When sharing data in a collaboration, the appropriate organization is critical. It is important, for instance, to have a coordinating unit or contact person, and there needs to be clear agreement on the various roles of the participants, the governance, and who makes the final decision on various issues. At least one person needs to be familiar with the policies and procedures for the different members of the collaboration. A “data steward” can be helpful, and researchers will know to approach them to answer questions.

Breakout discussion participants recognized that arrangements need to be made for the training/education about the project for anyone who comes in new and does not know where to start. That training can include an overview of the entire project, not just the parts in which the new person will be involved.

Furthermore, it is important for every effort to engage the right people. There will likely be many different people available, but some will be more suited to this sort of collaborative work than others.

Finally, it was pointed out that everyone who is involved in the project needs to receive credit and attribution for their contribution. For instance, those doing the data work should receive the appropriate professional recognition. This sort of acknowledgment is particularly important in cross-sector collaborations, where it is easier to overlook the contributions of individuals whose work is less known. A crucial prerequisite for appropriate credit in a research-data consortium is transparency in the data provenance, credit, and use policies.

EXAMPLES OF PRINCIPLES FROM OTHER GROUPS

For background and for inspiration, summit participants were provided with two sets of principles related to scientific data and research that had been developed by two previous groups. The first was a set of principles

adopted by a coalition of cloud storage providers, while the second was a collection of principles for scientific publishing from the International Science Council (ISC; additional details below).

Trusted Cloud Principles

In 2021, a group of cloud storage providers including Amazon, Google, Microsoft, Cisco, and IBM announced that they had developed a set of principles they were committed to upholding in their provision of cloud storage services. The purpose of the principles, according to this group, was to “maintain consistent human rights standards across their servers while also ensuring that cloud services providers’ interests are protected” (Arghire, 2021).

The following context for and listing of the Trusted Cloud Principles is taken from the official announcement for those principles:

Trusted Cloud Principles signatories (Amazon, Atlassian, Cisco, Google, IBM, Microsoft, Salesforce, SAP, Slack) are committed to protecting the rights of our customers. We have agreed to strong principles that ensure we compete while maintaining consistent human rights standards.

As cloud service providers we:

• Recognize the interest of governments around the world in protecting the safety, security, privacy, and economic vitality of individuals and organizations that use global cloud services;
• Recognize that international human rights law enshrines a right to privacy;
• Recognize the importance of customer trust and customers’ control and security of their data, which entails both safeguarding the data customers own in the cloud, and creating products and policies that establish, maintain, and enhance that trust;
• Support laws that allow governments to request data through a transparent process that abides by internationally-recognized rule of law and human rights standards;
• Support international legal frameworks to resolve conflicting laws related to data access, privacy, and sovereignty;
• Support improved rules and regulations at the national and international levels that protect the safety, privacy, and security of cloud customers and their ownership of data;

• Recognize the importance of publishing, on a regular basis, transparency reports detailing aggregate statistics regarding government data requests.

To achieve our objectives, we commit to working with the tech sector, public interest groups, and policymakers around the world, particularly in the countries where we operate or plan to operate data centers and cloud infrastructure, to ensure that laws and policies are substantially in line with the following principles:

1. Governments should engage customers first, with only narrow exceptions. Governments should seek data directly from enterprise customers rather than cloud service providers, other than in exceptional circumstances.
2. Customers should have a right to notice. Where governments seek to access customer data directly from cloud service providers, customers of those cloud service providers should have a right to advance notice of government access to their data, which only can be delayed in exceptional circumstances;
3. Cloud providers should have a right to protect customers’ interests. There should be a clear process for cloud service providers to challenge government access requests for customers’ data, including notifying relevant data protection authorities;
4. Governments should address conflicts of law. Governments should create mechanisms to raise and resolve conflicts with each other such that cloud service providers’ legal compliance in one country does not amount to a violation of law in another; and
5. Governments should support cross-border data flows. Governments should support the cross-border flow of data as an engine of innovation, efficiency, and security, and avoid data residency requirements (Trusted Cloud Principles, 2022).

International Science Council

In 2019, members of the International Science Council were asked to identify key issues for science, and scientific publishing was identified as perhaps the most important scientific policy issue. This led to the formation of an international working group, which in turn developed a set of principles for scientific publishing along with an assessment of the need for

reform. The eight guiding principles they identified for improving scientific publishing in the digital era were endorsed by ISC members through a resolution adopted at the 2021 ISC General Assembly (ISC, n.d.).

Those eight guiding principles were the following:

1. There should be universal open access to the record of science, both for authors and readers, with no barriers to participation, in particular those based on ability to pay, institutional privilege, language or geography.
2. Scientific publications should carry open licenses that permit reuse and text and data mining.
3. Rigorous, timely and ongoing peer review must continue to play a key role in creating and maintaining the public record of science.
4. The data and observations on which a published truth claim is based should be concurrently accessible to scrutiny and supported by necessary metadata.
5. The record of science should be maintained in such a way as to ensure open access by future generations.
6. Publication traditions and bibliodiversity of different disciplines and regions should be respected.
7. Publication systems should be designed to continually adapt to new opportunities for beneficial change rather than embedding inflexible systems that inhibit change.
8. Governance of the processes of dissemination of scientific knowledge should be accountable to the scientific community (ISC, n.d.).

These key principles along with some explanation and elaboration can be found in the paper “Key Principles for Scientific Publishing: And the Extent to Which They Are Observed” (ISC, 2021). A second paper, “The Case for Reform of Scientific Publishing,” makes suggestions for how scientific publishing might be reformed in accordance with the eight guiding principles (ISC, 2023).

PRINCIPLES FOR RESEARCH DATA COLLABORATIONS

After discussing principles in the breakout groups and returning to have a discussion with the full group, summit participants identified principles for research data collaborations, which could be adapted and put into use by research data organizations. There is some unavoidable

overlap among these principles as they are interconnected and reinforce one another.

1. Empower and Recognize Data Workers. Implement professional recognition programs and provide ongoing training for individuals involved in data work, with special attention to data provenance and potential contradictions among principles.
2. Foster Trust, Transparency, and Reciprocity. Establish transparent data practices, regularly communicate with stakeholders, ensure reciprocal relationships in data exchanges, and place a strong emphasis on data provenance to enhance trust.
3. Base Decisions on Evidence and Foster Open Communication. Use evidence to inform decisions, promote open communication by minimizing jargon, agree upon standardized terminology, and pay special attention to data provenance for ensuring evidence integrity.
4. Build Sustainable Infrastructure and Ensure Compliance. Develop and maintain data infrastructure that is environmentally and economically sustainable, while adhering to legal and regulatory requirements. Ensure compliance with a focus on data provenance and address any potential contradictions with sustainability practices.
5. Promote Fairness in Collaboration. Allocate costs and recognition in data collaborations fairly, ensuring representation and benefits a wide range of populations. Identify potential contradictions in fairness considerations and address them transparently.

Empower and Recognize Data Workers

Data workers form the core of any cross-sectoral data-research collaboration, and breakout discussion participants recognized the importance of empowering and recognizing them. Collaborations need to implement professional recognition programs for the people carrying out the data work and also provide ongoing training for these individuals. It would be valuable to institute education on data work as early as K–12, with attention to principles and ethics. Data stewardship needs to be professionalized, with the collaboration members paying for needed work and training.

Breakout discussion participants also pointed out the need to focus on data provenance and credit, as these are particularly important in collaborations with data coming from multiple sources.

There is a potential conflict between recognizing individual contributions (which may emphasize specialization) and the need for collaborative efforts (which may require a broader skill set). Professionalizing data stewardship may also conflict with the desire to involve people from lower-resourced communities that provided data in working with those data.

Foster Trust, Transparency, and Reciprocity

Collaborations work best when there is trust and reciprocity among the different participants. Trust is most easily established and maintained with transparency. This principle calls for establishing transparent data practices, regularly communicating with stakeholders, ensuring reciprocal relationships in data exchanges, and placing a strong emphasis on data provenance to enhance trust. Data are most useful when they are FAIR (findable, accessible, interoperable, reusable), trustworthy (with proper documentation), and maintained on sustainable platforms to guarantee long-term accessibility.

One possible contradiction among these various principles is that transparency and reciprocity may be at odds in certain situations. For example, complete transparency may not always be achievable or may conflict with privacy concerns.

Base Decisions on Evidence and Foster Open Communication

Basing decisions on the best available evidence is also an important component of successful collaborations. Research-data consortiums need to use as much reliable evidence as possible when making decisions. Special attention can be paid to data provenance in order to ensure the integrity of the evidence.

Decision-making in a cross-sectoral consortium also needs to follow other principles. To begin with, for instance, there needs to be agreement on the problem that people are attempting to solve with data. Both the problem and the specific question being asked need to be clearly specified. Open communication is key to making sure that there is indeed agreement on the problem being solved. Such open communication can be promoted by minimizing jargon and agreeing upon standardized terminology. Conflict may arise between the need for evidence-based decision-making and the urgency of particular situations where timely action is required but comprehensive evidence may not be readily available.

Build Sustainable Infrastructure and Ensure Compliance

Cross-sectoral data-research collaborations generally require a significant commitment of time, money, and expertise, but they have the potential to pay valuable dividends for many years to come if they are built to last. It is more likely that these benefits will be realized if the policies, technology, and other aspects of the data infrastructure are developed and maintained in a way that is environmentally and economically sustainable, while also adhering to all legal and regulatory requirements.

Sustainability requires growth, adaptation, and learning. Therefore, a research-data collaboration can engage in experimentation on data use, applying the lessons learned, and sharing those lessons inside and outside the collaboration. A collaboration will have both technical and social dimensions. It is important to distinguish between the two and to maintain the health of both types. Within the collaboration having clearly identified roles and accountability will make it possible to sustain the project’s health.

Maintaining compliance and accountability is also vital, and members of the collaboration need to be pragmatic about costs as well as about legal and regulatory issues. The privacy of the data will always be paramount. Cutting corners in these areas may seem to have short-term benefits but in the long term will threaten the sustainability of the project.

One conflict among these principles may arise in balancing the sustainability of the data infrastructure with the need to comply with rapidly evolving legal and regulatory requirements, which may require significant resources.

Promote Fairness in Collaboration

Ensuring fairness is important in any collaboration, but it carries meaning in research-data collaborations because of the treatment that many minority populations have historically received relative to research data collected from those populations. The Tuskegee syphilis study is the classic example of this, where nearly 400 Black men with syphilis were enrolled in a study without informed consent and were never given the treatment that could have cured their disease (Tobin, 2022). A more recent example was work done on blood samples collected from the Havasupai tribe, also without their consent (Pacheco et al., 2013). A vital principle in research-data collaborations is that that the costs of and credit for these collaborations need to be allocated equitably, with representation and benefits being pro-

vided to participants including members of historically underrepresented populations.

To uphold this principle, it is important that data are collected, assembled, and used in cross-sector research-data collaborations following the FAIR data principles, and the process of organizing and operating the collaboration needs to follow the JEDI principles so that the collaboration represents and benefits the populations it works with and is intended to serve. Beyond that, particularly with data collected from Indigenous populations, the collaboration needs to follow the CARE (Collective benefit, Authority to control, Responsibility, and Ethics) Principles for Indigenous Data Governance (GIDA, n.d.). In contrast with the FAIR principles, which are mainly concerned with ensuring data sharing, the CARE principles consider power differentials and historical contexts when considering the application and use of data. Both the JEDI and CARE principles emphasize bringing affected populations into the conversation about how data can be used.

A potential challenge to promoting fairness in a collaboration may arise when trying to balance the need for equitable resource allocation with the reality of resource scarcity. This could be particularly tricky when serving diverse populations with differing needs. It is important to identify potential contradictions in fairness considerations and address them transparently.

POTENTIAL PITFALLS

The participants in the breakout sessions identified several potential pitfalls in establishing cross-sectoral research data collaborations. To begin with, while the principles may be straightforward to identify, and groups may “adopt” them with relative ease, they often prove difficult to put into practice. As one participant commented, “We are drowning in principles. We lack [the] ability to get past the principles. They are easy to draft but hard to implement.” A collaboration may set forth policies based on the above principles but find, for instance, that implementing standardized terminology, practices, and training is much harder in practice than in theory.

Another major barrier can be the tensions between sectors and the differences in incentives and culture that make it difficult not just to understand the acronyms and the incentives from other sectors, but also to actually have a conversation around what the purpose of a collaboration might be. Somehow people need to be able to understand the cultures of the sectors they are working with, including the barriers that different

cultures—especially the dominant culture—may create that keep people from participating. A further potential stumbling block is the “I’m special” attitude. Each domain is most comfortable with its specific way of doing things, sees that way as best, and wants to do things its own way. When everyone wants to be “special,” interoperability becomes limited.

There can also be tension between various principles and practices. For example, there is much to gain from professionalizing data stewardship, but at the same time it can make it difficult to empower everyone who wants to participate. That is, if a research data collaboration is going to pay trained individuals to be data stewards, it can disenfranchise those who do not have the ability to get that training. How can one both make data stewardship a profession and empower lower-resourced groups to participate?

Summit participants also identified legal and financial pitfalls. For example, intellectual property can prove to be a particularly complex issue in cross-sector organizations. It can also be extremely difficult to allocate compensation and cost-bearing fairly. How does one account for lost opportunity? Who bears the financial burden of the data sharing? And figuring out incentives can be challenging. All of this is made more difficult because of the culture shift involved when becoming part of a cross-sectoral collaboration.

REFERENCES