Viewers Choice Critique

Link to Website/visualization: https://projects.interacta.io/country-tsne/

Purpose of the interactive website?

The purpose of this website is to help inform and tell the story of how similar or unique countries are given the varying factors which describe them. This site uses a machine learning algorithm to put the different countries into their own clusters based on multiple categories.

We all know there are certain characteristics which group certain countries together. Whether it's population, size, gross domestic product (GDP), or even more unique measurements such as government spending and civil liberties scoring, there are numerous ways for us to show which countries are more like others. But there are ways we can find out how countries differ from each other (example: USA versus Spain) while also discovering how alike two random countries could be (example: India and Argentina).

Overall, this visualization can teach us more about what defines countries and give us an outlook of what areas countries need to improve in. Along with that, we can find out who a countries "neighbors" (similar countries) are when we toggle between a number of different factors.

After clicking the "Run" button at the beginning of the website, we get to the main webpage which shows us the groupings of similar countries.

Each country is represented by a circle and the size of the circle depends on the number of the factor in which you are invoking. The default factor which starts is population and the color of each circle depends on which continent it is located in. Both the size and the color can be adjusted based on a list of choices, which can be accessed by clicking on their respective drop-down menus. For example, if we were to change the color scheme to based on surface area, the colors would be different.

When it comes to how the countries are clustered, if you click on the right, it will show a list of properties (as shown by the screenshot below).

Initially the algorithm has a certain number of factors already checked. This is explains why certain countries are grouped together because they have similar scores in those given factors. The user has the ability to check and uncheck each category as they please. By checking and unchecking certain categories, the clustering will then be alternated and changed. This way we can see which countries are similar based on the factors or categories we choose. There is also a reset button to change it back to the original settings.

If the user wants to look at an individual properties of a country and not just the world as a whole, they can click on any bubble. A list of sliders will show up with all the attributes of a country (shown below) and some of these values can be adjusted. Properties such as population and surface area can't be changed while others like GDP and unemployment rate can be adjusted via the scale. As you change those values, you will see the that given country shift and be grouped into with other places which have similar scores to those which the user adjusted.

Where did the data come from and what is it?

The data for this visualization is public and can be either downloaded or view through an Excel spreadsheet. It's available to view through this link. It's named as the WDVP Dataset and has a list of countries along with scores for a number of indicators. From population size to financial freedom score, there are values for nearly every country. Along with that, there is a row which tells notes on what the indicator is measuring and how it is scored.

Also, under the name of each indicator is a source name along with a website URL of where it came from. For example, the data for number of people in each given country was given by the World Bank Group and a direct link to the dataset is provided. This way, it specifically shows how the data was collected for this project.

Below is a screenshot of how the dataset looks like when it's first opened

Who is the ideal audience for this visualization?

Due to the visualization containing a vast array of information regarding numerous countries, it could be that there is a wide audience who this website was made for. Most likely the people who will be looking at the data and visualizations shown on the website will be government workers. People who are studying or working in the political science field fall under the realm of being part of the ideal audience for this visualization as well. By using the data displayed, this visualization can help create show how certain scenarios would play out and that can help when it comes to political science research. It answer questions researchers and authors have while also creating new ones, which sparks debates along with making topics for papers and talks in the future.

This visualization is for anyone who has an interest in global politics and seeing the way the world is organized.

What kind of questions are asked by this?

Often in research papers, the authors want to suggest potential solutions or explain that if a certain trend in a country or region keeps up there will be consequences are due to it. This visualization can help with that. For example, if someone was studying the country of Indonesia, by looking at the initial factors set by the algorithm, they can see that Indonesia is in the same grouping as countries such as Colombia, Panama, Philippines, and Mexico (shown below). We can now ask questions such as:

• How would it be if Indonesia increased their GDP and explore which countries they would become similar to
• If we added other measurements such as education expenditure per person, how would that change the way countries are grouped?
• In addition how would the clusters be formed if a factor from the original algorithm was taken out, such as government spending?
• What would the impact be on countries in a certain continent (for example: Europe), if all of their GDP rates were raised?

We can do this for individual countries or even analyze certain regions/continents if we liked. We can also re-shape the data based on one of the factors shown above previously. We can envision scripted scenarios when it come to the development of certain countries and see what happens.

How can the website help us answer this?

By using the tools given on the right hand side and at the top of the website, we can get some more clearer answers to our questions we asked above. For example, if we adjusted the GDP of Indonesia by a huge amount, we can see that it shifts the neighbors that Indonesia has in this visualization. They begin to sway more away towards more of the bigger countries such as the United States, China, and India .

A positive about this visualization is that when you do adjust the slider on a country's setting, it gives a white triangle to show where the slider originally was. This is in the case of the user wanting to put an individual slider back to its original value without having to reset all of the countries settings.

Let's say we wanted to see which countries are similar when it comes to education. We can adjust the settings of the visualization to display just that. There are three settings which fall under the category of education. They are: education expenditure, expenditure per person, and school life expectancy. By changing the settings to only account for the ones we mentioned, the map turns out to look like this

By only applying a limited amount of settings we have more clearly defined clusters of countries and it's easy to distinguish between the groups. We can still pick out the major countries as the size of the individual circles are being defined by population. Here we can answer questions such as which countries in a certain continent are grouped together and which places are more similar to the United States when it comes to education.

What works well?

In terms of positives for this website, I think it does a good job in terms of covering a broad topic and coming up with a unique approach to it. It displays the data in a way which entertains the user. It gets them curious as to how the visualization works and why the initial groupings are set up. The website encourages interactivity by allowing the user to change how the graphs are made and can cause some monumental shifts in the display. The full interactivity is a plus as it allows the user to explore every possible option given by the dataset. This gives the user full power to change not only the representation of the data but also to tweak certain parts of the data to show potential changes.

Animations are also a big help in this visualization as the data points are moving depending on the changing of input. At times it can get a little bit wild if you drastically change the dataset but overall, the site does a good job of the movement of the countries. There is also a search bar at the top right where you type in a country and it will immediately take you to where the it is located on the chart. Sometimes the display can be a bit too much with so many countries shifting all over the place so it's a good resource if someone is researching just one particular place.

Another positive is how the website has a little helper at the bottom which can take you step by step through the process of how to find the data, how to change it, and it adjust the graphs accordingly.

What needs improvement?

There are a couple of parts on this visualization which I would suggest need improvement.

The first will be the that the data isn't all that clear to begin with. Although it does give a walkthrough on how to navigate through the data, without it and at the beginning, the visualization can be confusing to look at. It's tough to first figure out why certain countries are grouped with others so the information really doesn't pop out initially. Along with that, it doesn't explain the definitions in each of the measuring points and these things can get quite complex in their measurements. Not everyone is going to know what health expenditure is going to mean when they read this so a bit of an about page or an index explaining what each of these factors are would have been helpful. That way it would make the user feel more comfortable when navigating through the site.

Another issue is the shift in zooming in and out with the visualization. Since there is a lot of countries, there are a lot of data points and if you adjust it, you could get an entirely different graph. While looking through this website, I found it tough to keep zooming out whenever I made small changes to the data. Because certain points would go out of the window frame and it would take a long time to zoom back out and see the whole picture.

When it comes to color, I thought that having bright colors for each of the countries made them stand out pretty well so the names are pretty easy to read out. Same thing goes for each of the settings in that they are pretty easy to read out. But the background color could be changed. In my opinion, they should have used a more brighter color as the background and not gray. It sort of dampens the overall view of the data and I felt like a different color choice could have made it stand out more.

One final critique would be that during when running of the algorithm, the user can see the data points move and not always are the same points in the same region each time you run it. The initial groups are always similar but not the location, this can be a bit confusing at times for users if they reload the webpage or visit it at a different time. Unless they have the tab open, they will get different points in different places. Also when the algorithm is still running, the data points are available to interact with and the points are still moving. It can be a bit tedious for the user to click on the points that are moving and you have to wait for them to fully stop to see what the groups are. Otherwise the user can interpret the data in the wrong way if they see the points move, click on a different tab, and see the finished product when they click back to the visualization. It can cause some confusion for the user. Like in the screenshots below, after running the visualization two separate times, you get two different maps of the data points.