Despite best intentions, there are countless ways an organization’s data management can cause harm to individuals and groups of people. For example, institutions collect data that they think will help them make faster and more precise decisions with data, but the use of those data end up perpetuating exclusionary practices, like credit scores or the use of the geographic location of past criminal events as a deployment strategy for police surveillance. In both these cases, the data ensure further and repeated negative actions are taken against individuals and neighborhoods already targeted for “remediation” or “increased intervention” nearly ensuring those people don’t exit the doom loop of the algorithm.
Many government institutions still treat gender as binary. Even when systems are updated to be inclusive of more than two genders, there are not always the right policies in place to allow non-binary individuals to review and modify their records. And this is only one instance of many where a person may rightfully need to change their data records.
Even the name of a data set can cause harm. Many education institutions track student disciplinary incidents under the term “student behavior data” or a similar naming convention. But research reveals bias in the adult administration of student disciplinary actions (detentions, suspensions, etc.). Not only should efforts be taken to reduce the bias in those actions, but the entire naming convention should change. To use the name “student behavior” in the analysis and interpretation of those data prime the audience to instead think of those data as “fact” about the students on which it reports. Those data are really tracking adult behaviors in the classroom and school.
In each of these instances, I doubt the data managers and analysts sought to cause harm. So how can public sector data teams avoid harm in the future? Strengthening data governance, a practice which already promotes the ideas of shared data management decision making across many stakeholders and democratization of the data collected within the organization, offers an opportunity for organizations to be even more intentional and inclusive in that governance work.
Below I provide a few examples of how this idea of shared decision making and shared data access can be amplified. But first I want to take a moment to discuss the concept of equity, of which there are many frameworks that could provide insights on improved data governance practices. For the purposes of this article, I chose the Design Justice Principles, created by the Design Justice Network and spelled out in wonderful detail in the book Design Justice by Sasha Costanza-Chock. The Design Justice Principles total ten, but I will highlight three of them.
Principle 2: We center the voices of those who are directly impacted by the outcomes of the design process.
Design Justice Network
Principle 4: We view change as emergent from an accountable, accessible, and collaborative process, rather than as a point at the end of a process.
Principle 6: We believe that everyone is an expert based on their own lived experience and that we all have unique and brilliant contributions to bring to a design process.
Note that the principles are developed around the concept of design, particularly of new data products. But the principles can apply to governance as well, as both involve decision making on behalf of a greater whole. So how might we use these principles to support a better and more equitable data governance practice?
Understand and Leverage the Perspectives of Your Data Governance Committee
First, leverage your organization’s existing data governance committee. (And if you don’t have a committee, start by forming one!) Who sits on your data governance committee? These members should represent a cross sample of programmatic data functions across your organization and should be empowered to make decisions on the management of the data discussed. Typically these folks serve at a director level within the organization, having official ownership over certain data sets. But in the data governance committee meetings, the members also bring perspectives beyond their work role, based on their own identities: race, ethnicity, gender, religion, marital status, age, income, ability, even where and how they grew up. It is important to understand how those perspectives enhance or limit your decision making, and to seek a demographically diverse team.
That said, I realize their are human resource considerations that may extend beyond the purview of the leaders tapped to establish those committees. Who is recruited, hired, mentored, and promoted are entangled in complex social and policy systems that will take time to unpack and rebuild in more equitable ways. And even under perfect human resource circumstances, you will likely not be able to fill every gap in your committee’s understanding as they create policies on behalf of the communities they serve. This is why it is important to couple the data governance committee’s efforts with community input on key decisions.
Recruit a Community Representative Group (or Several!)
Working with your community members who represent the people most affected by your data decision making is an excellent extra step to protect your organization’s constituents from unintended harm. This work could mean tapping into the expertise of existing community groups or forming groups specific to the data governance task at hand. And the duration of each group’s support could be temporary or ongoing. The key in forming or recruiting these groups is to:
- Be intentional about bringing diverse perspective to the table, not just folks who will agree with you
- Center the voices and opinions of the community representatives, not the data governance committee, during discussions
- Share information readily with the community group in support of their deliberation and decision-making activities, creating transparency and shared understanding
- Compensate the members for their time and expertise
- Ensure that the role of the community group and its decision making parameters are clearly articulated to that group’s members and that the power to make decisions on key data governance matters is appropriately shared between the data governance committee and community group
This last point means these community groups are not just brought together to give advice but have real power in affecting outcomes. The idea of power sharing is usually the hardest one to implement. In some cases, the organization is legally obligated to take certain actions and thus cannot turn over full power to an outside entity. And even when power can be shared, it is human nature to not want to hand it over out of worry that “outsiders” simply can’t understand the difficulty of the decisions at hand. Also, people have egos. It takes a real shift in individual and institutional mindsets to effectively work together. Each collaboration between the community and the data governance committee should be examined individually with its own set of roles and rules on how the entities will work together. Sherry Arnstein’s 1969 Ladder of Citizen Participation is a helpful starting point to understand the importance of those shifts in power and how to move toward greater community involvement in critical data decisions.
The ways in which you can use a community representative group are numerous, and organizations don’t always have to rely on the same group to support the data decisions the governance committee will tackle. Some ways to engage the community group include:
- Development and prioritization of data use cases for future data collection and data system design
- Creation or review of key data policies, particularly around data collection, data privacy, informational self determination, and data sharing
- Approval of data sharing requests that have a higher probability of causing harm to certain groups of people
Information + Transparency = Better Data Stewardship By All
The data a public sector organization collect should be considered a public asset, and thus government agencies should make every reasonable effort to safely share as much of the data it collects as possible (both internally to other employees of the organization and to the public) as long as it benefits the public interest and meets privacy standards. This data sharing may occur through open data programs and sites (where appropriate), but for more sensitive data, the sharing will take place on a case-by-case basis through a robust data sharing policy and review process.
Most public sector organizations have some data sharing policies and processes in place. Some are better than others. But if there is a common place where all government agencies could improve, it is in providing an adequate level of context for the data it shares so that it can be analyzed and interpreted appropriately. Data governance committees can mitigate this harm by spearheading efforts to create greater transparency on the origins, attributes, and definition of those data. Key information on the source of the data collected, the intent behind collecting it, its formal definition, and even the history of the data collection effort can help prevent accidental misuse.
You don’t need mature data governance structures to start deploying these ideas. Heck, you don’t even have to have a data governance committee in place. You can use these ideas as a place to start or grow your data governance practice. (And if you don’t know the first thing about where to start, here is a shameless plug for my University of California Berkeley online course Data Governance for Public Decision Making that provides the basics on setting up a sustainable practice.) Now is always the right time to start tailoring whatever data governance structures you have to promote better equity in your data management decision making.
Shift Forward Consulting LLC 