Building a Data-Driven World Map
How we designed an alternative data-driven world map and received the Grand Prix from the Prince of the UAE
World Data Visualization Prize
In February of 2019 in Dubai, we received the main prize in the World Data Visualization competition. The crowned prince of the United Arab Emirates personally awarded the Prize, and the contest itself was held in partnership with the World Government Summit.
The Summit chose DataVis for one of its special awards. On the one hand, it highlights the relevance of this area, on the other — it raises its status. Government Summit that usually deals with the most important global issues has driven attention to how to get the most out of working with data. The organizers invited David McCandless, a well-known data journalist and one of the best experts on data visualization, to curate the entire project from collecting applications to appointing the jury. It was a great honor for us that it is under his leadership the jury awarded our project.
Jury members: Stefanie Posavec — Dear Data, Alberto Cairo — the author of The Truthful Art and The Functional Art, Patrick Burgoyne — the editor of the Creative Review design magazine, Alexandra Mousavizadeh — the director of Legatum Prosperity Index, David McCandless — the founder of Information is Beautiful.
In the call for applications, David McCandless wittily remarked that now participants will have something to do on holidays. That’s exactly how it happened: we prepared the project during the New Year holidays. At the end of 2018, the organizers proposed three questions that in different ways develop the general topic — “How governments are improving citizens’ lives”.
The summit set the concept and provided datasets, participants created visualizations that told stories. One of the objectives was to present beautiful solutions to the highest guests of Dubai and maybe stimulate their interest.
Alternative World Map
The first days of January are the best to think about improving living standards around the world. We chose the question “What makes a “Good” government?”, but did not answer it. It was more challenging to suggest a tool that will help everyone find an answer. And you can try it too 👉.
The question came along with a small dataset, which described each country with 32 measurable parameters ranging from population, size, happiness index, GDP, level of corruption and number of women in government. Despite that the organizers called to think about the measures for success of various governments or look for patterns in figures, only economists or politicians can treat such data professionally. Although we can help them with it.
We have created an alternative world map and reflected what will happen to the situation in the country if certain parameters change. Since we do data visualization, and not geopolitics, we refrained from evaluating this data in any way. We employed a t-SNE machine learning algorithm to process data, and you are left to play with the unemployment rate, health care costs and other characteristics.
How is the alternative world map organized
Each point on the map stands for a country, points are grouped not according to geographic proximity, as on a regular map, but by a number of characteristics.
You can customize the display mode: set up the criteria for dividing points by color and size, manage the parameters for the algorithm to consider.
The map turned out to be magical:
- It merged an analytical algorithm and visualization. The project’s design allows to impact the operation of the algorithm. This reinforces the general idea of data visualization: design not only beautifies but also makes work with an array of information convenient and fast.
- It is fully interactive. In the list you can choose what will the color or size of points indicate in order to explore the map effectively. On the separate panel on the right you can experiment, turning on and off different properties or vary their values, and that will affect the outcome.
- Our tool allows to experiment very quickly. Just try to improve life in a separate country by raising the level of health care and GDP, and see how its position on the map changes.
We did not set ourselves any analytical or evaluative tasks. However, inevitably this DataVis encourages to make observations. For example, it is easy to notice that so-called “developed” countries immediately join the same cluster. Some countries became neighbors, though generally they rarely correlate, for example, the Middle East and Singapore, or Russia and Brazil. Our map brought them together after having analyzed a meaningful dataset. In order to help study the map more informally, we educed a few examples of how it works. They show how to figure out what a country needs to change in order to acquire its “alternative geographical position”.
How does the alternative world map work?
The map is powered by a machine learning algorithm called t-SNE. Its main goal is to analyze tens of characteristics and identify local and global similarities. It is important to understand that t-SNE treats data utterly objectively: it does not care what it is about.
At the same time, it is arranged nonlinearly and can adapt to changes. This means that the initial position of points and the algorithm’s operation time will affect the result as updating. When the latter has reached a balance and holds it for some time, it will be “more difficult” to adapt to data changes, since they will be evaluated in the new circumstances.
The algorithm already has a built-in ability to visualize. Since we developed a functional and elegant decision-making tool, we wanted to make the design recognizable, pleasant in use and utterly clear. Therefore, it looks like a world map (and a bit like a starry sky).
Thus, we transferred the dataset to the t-SNE algorithm, created our own design, adjusted the interaction between the visual interface and the algorithm. And these elements do not obstruct each other.
What can come out of it
Alternative world maps can become a tool for decision-making. We believe that one of the data visualization missions is to tell a long and complex story concisely and clearly.
The map is universal and can work with any data on countries. We worked with the data provided to us in the contest. Its sources are listed in the original file. And although data objectivity is one of the biggest modern myths, still due to visualization one can observe trends and patterns that are completely invisible in the tables.
«This dazzling, winning entry from Nikita Rokotyan uses artificial intelligence and machine learning to find previously-unseen connections and harmonies between different countries. It uses an AI technique called t-SNE to discover clusters of nations that are related by happiness score, health expenditure, investment in education and many other variables. It then presents these patterns visually, creating an interactive world map that we can explore, tweak and filter to find unexpected pairings and insights. It uses design, code and artificial intelligence to bring data and statistics to life», — David McCandless, https://informationisbeautiful.net/2019/winners-of-the-world-data-visualization-prize
Team
Nikita Rokotyan, Olga Stukova, Daria Kolmakova
