Have you purchased a product after seeing a social media advertisement that was closely aligned to your preferences? Or have you perhaps tried any of those smart applications that can tell you how you will look when you get older? They rely on predictive analytics models that make use of past data to predict future behaviour. 

Looking back, looking ahead: Transforming data into future insights  

Although predictive models are everywhere, they are not new. The first known models appeared when humans started growing crops, deciding when to plant, plow, and harvest. Puzzled by the movement of the stars, humans started to build models to explain and predict different phenomena, planet orbits, comets, etc. Some of these models generated accurate predictions, even if they were based on wrong premises. The Ptolemaic geocentric model was used until the 16th century to predict the location of the stars. This was done with the understanding of the universe as a set of nested spheres surrounding the Earth.  

Fast forward to the 21th century! Data availability has exploded in the last few years due to the internet and the big data revolution. Along with this increase, predictive analytics applications have become ubiquitous in many fields. In the UN we have a strong record of applying predictive analytics in many different critical areas, including to assess the impact and future risks of climate change, demographic prospects, humanitarian response and forecasts for the global economy.  

Understanding the benefits and risks of predictive models  

There are many benefits of using predictive analytics. It offers the possibility to allocate resources in more effective and efficient ways, reveal patterns and trends that are invisible to the human eye in large amounts of data, perform complex analysis using structured and unstructured data, and respond to problems proactively rather than reactively.  Predictive Analytics can help us facilitate data-evidence decision-making in very diverse situations. In most instances this contributes to improving operational excellence in the workplace.    

However, there are also inherent risks in the use of predictive analytics. Some of the most important risks are related to potential breaches in data privacy, the use of poor quality and biased data, and the application of biased algorithms, which can empower unethical practices and lead to discriminatory results and unfair treatment.  

The challenge with data privacy is more complex through predictive analytics since sensitive information about individuals or groups can be predicted, potentially without the data subjects' knowledge. This can include anything from less sensitive or more readily available information (proxy data) for instance, on social media channels.  

There are multiple ‘decision points’ within the design and implementation of a predictive analytics model. Sometimes human prejudices and bias can affect accuracy and efficacy. Bias can be intentional or unintentional. Algorithms are expected to act without any of the biases or prejudices that affect human decision-making. However, an algorithm can unintentionally learn bias from a variety of different sources. For instance, the data sets used to train the algorithm, or even the staff developing the algorithm can be sources of bias.  

In addition, predictive analytics not only has risks; it also has its limits. Some events can be challenging to prevent, and predictive analytics does not always give us an accurate answer about the future. This is called a “Black Swan event”, a metaphor often used in predictive analytics to describe the limits of forecasting science. It is also a reminder to consistently apply ethical good practices in the design and implementation of predictive models. 

Strategic alignment for successful initiatives 

The Data Strategy of the UN Secretary General (2020-2022) identifies data and analytics as an important and strategic asset for the UN system to advance in crucial socioeconomic areas worldwide, such as amplifying climate action, promoting gender equality, protecting human rights, advancing peace and security, and accelerating UN reform for greater impact on the ground.  It calls for the UN family to build the capabilities to ensure that we are all able to make data-driven decisions. 

The role of big data and technology have become critical in every sector. In the same way that many organizations today use predictive analytics in fields as diverse as finance, healthcare, aerospace and manufacturing, we at the United Nations can leverage the use of forecasting techniques for greater efficiency in our work for people and planet. 

Learning predictive analytics 

In this regard, the United Nations System Staff College (UNSSC) has developed different learning solutions in the area of data analytics for UN staff.  

Our latest offering, Predictive analytics in the UN Context, has been designed to enhance UN staff knowledge and capabilities in the area of predictive analytics, and the application of basic techniques in the development of predictive models.  It clarifies concepts and principles related to predictive analytics, describes with practical UN cases some of the main approaches in predictive modeling and some of the algorithms used for supervised and unsupervised machine learning. The course also explores risks, limits and ethical considerations in the design and development of predictive models.  

As the United Nations continues to make greater use of data analysis, it is our hope that more UN personnel will begin to learn and acquaint themselves with the potential of data analysis techniques such as machine learning.