Effective and Ethical Data Sharing at Scale

Understanding basic cooking

The data sharing food pyramid

The content in this cookbook is organized around five groups of key ingredients for effective data sharing initiatives:

1. Mechanisms for building and sustaining trust (carbohydrates): Any healthy diet starts with a foundation of trust among data sharing partners, data users, and other stakeholders.

2. Shared value and benefits (fruits and vegetables): When all stakeholders share in the value and benefits of data sharing, they get the necessary vitamins and minerals (like you’d find in fruits and vegetables) to reduce the risks of initiative-threatening “diseases.”

3. Dependable data and infrastructure (proteins): Just like healthy proteins, data and infrastructure occupy the center of the data sharing plate and must always be accompanied by shared value and benefits.

4. Mechanisms for supporting knowledge and strengthening skills (dairy): Dairy helps develop and strengthen human bones. Similarly, knowledge and skills support data sharing functioning and operations.

5. Flexibility and adaptability (healthy fats): No diet is balanced without a healthy dose of fats or—in this case—flexibility and adaptability, which facilitate storing energy for when it is most needed.

Mechanisms for building and sustaining trustworthiness (carbohydrates)

Trust is the most important ingredient for effective multiparty data sharing and, precisely like carbohydrates, it provides initiatives with the energy they need to function and achieve their objectives. Recent research highlights the robust quantitative evidence that greater trust is associated with increased data sharing and that the impact of trust is particularly significant where initial levels of trust are low. However, the same research also reveals that, to reach optimal levels of data sharing, increasing trust needs to be coupled with other strategies or actions, which is why trust must be combined with other ingredients.

Simply saying that trust is important does not help organizations or individuals working on data sharing initiatives to clearly lay out how or what they need to do. For this reason, and in order to foster more effective data sharing, this section focuses on understanding effective mechanisms for building and sustaining trustworthiness in data sharing initiatives. 

What are we making?

For data sharing initiatives to succeed, data partners need to build and maintain the trustworthiness of the initiative and embed core values and ethical considerations in its functioning. 

The trustworthiness of a data sharing initiative is the result of three key elements: 

1. Trust among partners, 
2. Trust from the general community, and 
3. Accountability measures. 

It follows that recipes for building and maintaining trust and for embedding values and ethical guidelines in data sharing initiatives can be grouped into three categories, which are described in the next section:

Shared value and benefits (fruits and vegetables)

Fruits and vegetables contain essential nutrients for people to thrive. Likewise, a sense that all partners are benefiting proportionately is essential to creating effective partnerships and reducing the risk of threats to the initiative. While humans can live for some time without fruits and vegetables, their dietary absence in the long run results in significant health problems. Unless organizations see the value in partnering for data sharing, they won’t stay at the table for long. Data sharing fails when the perceived value and benefits are unequally distributed for extended periods of time. When benefits consistently go to a few instead of many, incentives to participate decline.

Benefits encompass more than mere economic value. Data sharing partners can profit from an initiative in terms of cost reduction, reputation, knowledge, skills, or even access to data generated by others. Therefore, the emphasis should be not only on the monetary value of data sharing but on understanding the broader scope of added value of the initiative for data partners. Tools such as the Data Ecosystem Mapping Tool and Guide developed by the Open Data Institute in the context of the Microsoft Open Data Campaign can help analyze potential value that can be exchanged between actors within a data ecosystem. 

Sharing value in a fair way is already complicated in the context of bilateral data sharing between the public and private sector, as the literature suggests, because reconciling diverging interests and expectations can be hard, irrespective of the number of stakeholders involved. In the context of multiparty data sharing, it can become even more complex to balance competing interests.

Dependable data and infrastructure (protein)

High-quality data and infrastructure (defined as the software and hardware underpinning data exchange) lie at the heart of successful data sharing. They can be compared to proteins, as they are generally put at the center of the plate and seen as the main course of a meal.

What is less obvious is that it is not the quantity of data shared nor the technological features of the infrastructure that ultimately lead to the success of data sharing initiatives, but rather their dependability. 

Dependability encompasses two aspects of data and infrastructure in data sharing: the trustworthiness of these systems and the quality and appropriateness of the information created in terms of user needs.

• Trustworthy data and infrastructure are safe, unbiased, and regularly monitored. 
• Quality and appropriateness are linked to the security and interoperability of datasets. 

These two concepts reinforce each other and ensure that data partnerships result in dependable products that organizations can use to develop new applications, services, policies, and operations.

What are we making? 

Data sharing initiatives may take different approaches to ensure their data and infrastructure are dependable, but what they generally have in common is their emphasis on:

• Ensuring technical safeguards are in place.
• Embedding quality assurance mechanisms in data collection and data analysis.
• Adopting appropriate interoperability and data format approaches.

A stronger focus on one or another of these elements depends mainly on the sensitivity of the data at hand, the characteristics of the data themselves (i.e., in terms of variety of sources or granularity), the targeted use cases and objectives of the initiatives, and the resources available. Levels of trust between partners also influence the attention paid to technical safeguards and quality assurance in particular. 

For example, data aggregators such as the Humanitarian Data Exchange, which is an open-source and open-access web platform enabling humanitarian organizations to share data, spend more time and resources on ensuring interoperability of datasets, whereas initiatives such as Global Fishing Watch, which seek to inform government response, focus their efforts on ensuring data quality. This is because the former collects a wide variety of data from different sources whose interoperability is key to making the data usable, while the latter needs very precise data to create accurate maps of illegal fishing activities.

Mechanisms for supporting knowledge and skills strengthening (dairy)

A common characteristic of successful data sharing initiatives is their emphasis on knowledge, skills and capacity building. Sharing knowledge and skills strengthens data sharing agreements, just as the calcium in dairy products helps build and maintain strong bones.

It is widely recognized within the development community that it is not sufficient to make data available and expect organizations to access or use it. Instead, there is a consensus on the need to build data use skills among communities of users and on the importance of increasing data literacy to achieve sustained data use. 

Often, however, users are not alone in lacking data skills. The landscape analysis, for instance, pointed out that partners of successful data sharing initiatives are concerned about their own organization’s lack of data skills and the difficulty of finding and hiring the right people. Furthermore, the Global Partnership’s work has highlighted that data governance skills, for instance, are among the major obstacles to establishing public-private partnerships. 

Finally, it emerged from this work that limited attention is paid to building capacity at the community level. The landscape study indicates that almost none of the data sharing partnerships reviewed engage directly with the communities from which the data originate. Some models with limited engagement mechanisms included participation from civil society organizations and national representatives. However, these did not aim explicitly at building the data capacity of these communities.

What are we making? 

Successful data sharing requires building the skills of the user community alongside the skills of key stakeholders and partners.

Identifying the set of skills needed for a data sharing partnership to achieve its objectives and staffing it with individuals who possess these skills is crucial to achieving success. 

Furthermore, data sharing partnerships need to know their users very well and to understand their barriers to and requirements for accessing the data. This requires partnerships to focus on the capacity of external stakeholders who are expected to use the data. 

Finally, to make data sharing fairer and more sustainable requires strategies to increase the data confidence of the communities whose data are being used. However, examples of how this can be done in practice are limited, in part due to the high costs of large data literacy programs that are perceived as falling outside the scope of specific data sharing initiatives. 

To increase the capacity of external and internal stakeholders, approaches can range from producing formal training tools to establishing more informal mechanisms for transferring knowledge. A few successful recipes for building institutional capacity are provided below.

Flexibility and adaptability (healthy fats)

Data ecosystems in which the data sharing initiatives operate are never static, but rather continuously evolve, particularly from technological, stakeholder, and cultural perspectives. It follows that governance, architectural, and operational choices cannot be set in stone, and the possibility of adapting or changing must be envisaged at the outset. 

Financial and funding approaches also cannot be immutable. Most data sharing initiatives start with specific project funding but need to transition to something else at a later stage when the project’s funding ends. Changing the funding model can mean diversifying sources of cash or adopting subscription or fee-based models. Whichever approach is chosen, this transition from one funding approach to another remains a delicate moment that can break even very successful (from a user perspective) initiatives.

Healthy doses of flexibility and adaptability are therefore as important in data sharing as healthy fats are for human diets. Reasonable doses of flexibility and adaptability are critical for data sharing to succeed in the long term and for partners to remain involved. 

What are we making? 

Like trust, flexibility and adaptability are generic terms that do not help organizations to figure out what exactly they should be doing to ensure success in data sharing. More concretely, what data sharing initiatives should be paying attention to is how to embed flexibility in their governance, operations, and data architecture and how to identify suitable sustainable financial models that allow them to evolve over time.

Unfortunately, the development sector does not provide many examples of flexibility embedded in operations, governance, architecture, and financing. On the contrary, this area remains under-explored and represents a topic for which successful recipes are missing. 

This is not to say that no initiative has successfully included flexibility and adaptability,but rather that there is limited knowledge and research on how it can be done. Even successful initiatives have not explicitly theorized how to ensure the implementation of an evolutive approach. The sections below highlight a few findings and examples from the landscape analysis, but gaps remain concerning concrete steps for data sharing partners in this area.