Elevating Data Governance in the Public Sector

Celio Oliveira is at the forefront of driving significant changes in how public sector data is managed and utilized at Health Canada. In his forthcoming presentation at the CDAO Canada Public Sector event, he will discuss the critical role of data governance in transforming public service delivery

In his role, Celio is focused on advancing regulatory operations and enforcement at Health Canada by emphasizing the importance of data literacy among senior leaders from the data governance executive steering committee, as well as ensuring that data security measures are robust and effective. He advocates for using data as a strategic asset to inform and improve health policies and practices.

Celio's presentation at CDAO Canada will delve into how public sector organizations can transform their approach to data from being merely a resource to a strategic asset. By enhancing data literacy and implementing stringent data governance practices, Celio will share insights on how data can be used to make informed decisions that align with organizational missions and enhance service delivery to the public.

Check out the full agenda and register for the event here.

Why do you think elevating data governance is so important for the public sector today?

Celio Oliveira: Data forms the foundation of everything we do today. It is essential for nurturing intelligence and guiding all our actions. Currently, we collect data continuously, often without explicit intent, through various services and societal interactions. This data, however, is not always managed as it should be. We face multiple biases—there are at least twelve recognized types, all influencing data. Without investing time and resources in properly disaggregating and cleansing this data, and in understanding data ownership, we can inadvertently perpetuate these biases.

For instance, take machine learning algorithms designed to predict criminal recidivism. If these algorithms are trained using data from a context like the United States, where it is documented that the prison system disproportionately includes black and poor individuals, the resulting model could unfairly bias recidivism predictions against these groups. This does not reflect an inherent propensity among any racial or socioeconomic group but highlights a systemic bias within the data.

Effective data governance enables us to clean the data thoroughly, remove biases, and ensure that algorithms—essentially neutral machines—can operate on accurate, unbiased information. Thus, robust data governance is not just about managing data but about fostering fairness and accuracy in everything from AI applications to broader decision-making processes. This is why I advocate strongly for enhanced data governance practices to prevent future issues and ensure the integrity of our digital advancements.

Can you elaborate on the issue of bias in data, particularly the confusion between correlation and causation, and its implications for drawing conclusions?

Celio Oliveira: That's a great question, highlighting another significant gap: education. Unfortunately, many practitioners are not fully trained across the necessary spectrum of data science, which encompasses statistics, mathematics, societal knowledge, and subject matter expertise. Each of these disciplines is crucial, especially when handling diverse datasets like those from national defense or health systems, which require profoundly different knowledge bases.

Thus, the distinction between correlation and causation is critical. This understanding can be achieved through robust statistical analysis and a comprehensive grasp of the subject matter. Therefore, instead of relying solely on data-driven decision-making, I advocate for evidence-based decision-making. This approach bases conclusions and recommendations on solid evidence, ensuring that decisions are well-supported and truly reflective of reality.

How important is it to keep a human in the loop, especially with the increased use of automated systems and AI?

Celio Oliveira: It's extremely important to have human oversight in AI-driven systems, and it’s critical to understand that having just one person behind an algorithm is insufficient. Effective oversight requires a community, not just to empower decision-making but to ensure diverse perspectives and checks are in place. This community must continuously analyze both the algorithm and its outputs, verifying that everything functions as intended and remains relevant over time.

Another often overlooked aspect is the persistence of model performance after deployment. Many assume that an algorithm will maintain its initial performance indefinitely. However, this is not the case due to phenomena like data drift, where changes in the underlying data over time due to cultural shifts, global events, or collection methods can lead to what's known as model drift. This shift can significantly alter the algorithm’s performance.

Consequently, it's vital to regularly audit these systems to ensure their accuracy and control for false positives and negatives. This auditing shouldn't be incessant but should follow a scheduled plan that suits the organization’s needs—perhaps biennially, depending on the data sensitivity and application. Finally, the auditing process should involve multiple individuals to avoid biases that can occur if the task is left to a single analyst. This multi-person oversight helps ensure that the evaluations are thorough and that any changes in the algorithm's performance are promptly and accurately addressed.

What specific challenges do public sector organizations face regarding data governance, particularly in areas like health data?

Celio Oliveira: In the public sector, particularly within Health Canada where I am currently positioned, we encounter significant challenges related to the sensitive nature of health data. This data often includes private details about individuals’ diagnoses, necessitating rigorous protections to ensure confidentiality during dissemination. Before we can release any health data sets, they must be thoroughly anonymized or randomized, and encrypted during transmission to safeguard privacy.

One practical challenge is the disclosure of information. For example, ensuring that the health data of an individual who has visited a hospital remains confidential, yet also recognizing if the same individual was in a homeless shelter the previous day. These connections are vital for delivering comprehensive services but are complicated by the need to protect privacy.

Additionally, senior management must ensure that only the most qualified professionals handle this data instead of using the only ones available in your taskforce with minimal or insufficient data knowledge. The decision needs to be made because they possess the requisite expertise to manage such sensitive information responsibly. This requirement is crucial to maintain trust and ensure the integrity of the data analysis and recommendations.

Another significant issue is the public’s control over their data. Unlike in Europe, where individuals have more authority over their personal information, including the right to correct inaccuracies, Canada has not fully implemented such mechanisms. This lack of control can lead to mistrust in how data is managed and used, highlighting a critical need for greater transparency and public engagement.

Overall, the main challenge lies in balancing the need to protect individual privacy with the necessity of sharing information to provide effective services. Achieving this balance requires robust data governance frameworks that ensure data integrity and transparency, while also fostering public trust and confidence in how their information is handled.

How can public sector organizations best balance the need to make data accessible with the necessity of maintaining data security?

Celio Oliveira: First and foremost, ensuring that servers hosting sensitive data are located within Canada is crucial. This is not just about hosting preferences but a strategic measure to ensure data remains within the protective barriers of national cybersecurity protocols and legislation, making it easier to manage and regulate.

Secondly, we must implement robust mechanisms against cyber-attacks. This includes employing advanced malware protection and ensuring that system access is tightly controlled based on clear security credentials. The government operates on a 'right to know' basis, which means individuals must have a legitimate reason and the proper clearance to access certain data.

Another guiding principle is the FAIR principle, which stands for findable, accessible, interoperable, and reusable data. This principle sounds straightforward but implementing it is quite complex. For instance, ensuring data is interoperable involves significant challenges such as gaining consent from data owners, anonymizing data appropriately, and ensuring it doesn’t foster unfair competition or violate privacy.
For example, consider the ethical implications of sharing sensitive health data. We need to ensure that we do not inadvertently disclose information that an individual has chosen to keep private, such as a serious medical condition. Therefore, when making data sets open, it’s imperative that they are anonymized to prevent misuse or unintended harm.

Lastly, the security of data transfer and the mechanisms that determine access rights, such as security clearances and the right-to-know checks, must be rigorously maintained. These measures build the necessary credibility and confidence in our data governance processes, ensuring that data is not only protected against cyber threats but also used responsibly and ethically.

What advice would you give to data leaders in the public sector on transforming data from just a resource into a strategic asset?

Celio Oliveira: The foundation of transforming data into a strategic asset begins with enhancing data literacy, particularly among senior leaders. As leaders ascend higher in their roles, they often become less technical, which can impede their ability to make informed strategic decisions involving data. Therefore, increasing data literacy is crucial to empower them to contribute effectively to discussions and decisions regarding data usage.

Data leaders need to cultivate an understanding of how data can strategically benefit their operations and objectives. They must ask, 'What is the purpose of this data?' For instance, integrating financial data with health systems may seem unrelated, but understanding the connections can provide crucial insights, such as tracking economic factors that influence public health trends like COVID-19.

Furthermore, it’s essential to discern why and how different data sets, which may seem unrelated, can be utilized strategically. Just using data because it’s available or because current trends, like deploying chatbots such as GPT for customer service, promote such usage, isn’t sufficient. Leaders must critically evaluate the strategic intent behind employing these technologies. What value does a chatbot add to citizen interactions? Does it align with the organization’s mission and service goals?

In conclusion, public sector data leaders should focus on building a strong foundation of data literacy, understanding the strategic implications of data, and ensuring that their data initiatives align with their department's mission and broader governmental mandates. This approach will prevent deviation from core objectives and ensure that data serves as a true strategic asset."

Want to learn more?

Celio Oliveira is speaking at CDAO Canada Public Sector, kicking off on June 18th, 2024. Join him and many more world-class data and analytics leaders at CDAO Canada Public Sector to learn about the trends driving value creation in public sector data and analytics. Register to attend here.