Global Trust is a work of information visualization which attempts to explore the relations between indicators of economic development and the feeling of trust and confidence that people feel in different countries overtime.

Created by Yoav Luft for KTH course "Information Visualization", 2020.

In the initial discovery process I wanted to explore the question "Does immigration affects the values of a country". I started by exploring both data sets, World Values Survey (WVS) and Gapminder (GM). This initial exploration brought up two problems: (A) WVS has a lot of data, so comparing values of two countries will require dimensionality reduction of some sort; (B) GM has no data about immigration, only about refugees.

My second attempt at discovery was around the question "What influences the trust people give in each other and in public entities?". To answer that question, I've explored WVS dataset and found 6 questions regarding trust and 12 questions regarding confidence in organizations. I've decided to look at the following possible influences: economic issues, violence and corruption. I found 5 GM datasets that are relevant.

Next, I explored the datasets using Pandas, and plot several relationships. I've also considered the question of temporality: WVS data is divided to 4 year waves, while GM is annual. Fig 1. shows exploration of the impact of aggregating GM time series to chunks corresponding to waves.

After some research, I decided to approach the question of displaying temporal relationships by using a method inspired by Time-Curves

[Bach et al. 2015]

. But in my case I had to deal with holes in the data: For every country and WVS question, I had only two points of data, but for GM I had many. I decided to interpolate the data between points for which there is both WVS and GM data, and extrapolate based on GM for years outside WVS samples. Fig. 2 and 3 shows explorations of displaying data from two dimensions + time, and for using time-curves inspired design.

I can divide my learning into several aspects: Technically, I've learned to use elm-visualization, and I'm more strongly convinced now that Elm is the best thing for the web, ever, although it is still young and not as feature rich as other frameworks. On the visualization side, I had explored several ideas for connecting multi-dimensional data and time, but I would have liked to explore more. On the data side, I did not come into any conclusions. There seems to be, for example, some relation between corruption and lack of trust, but corruption didn't seem to influence confidence in the government. If I had the time, I would have wanted to explore more relationships, maybe through scatter-plot matrices, or user defined aggregations.

Bach B, Shi C, Heulot N, Madhyastha T, Grabowski T, Dragicevic P.

in Data. IEEE Trans Visual Comput Graphics. 2016 Jan 31;22(1):559–68.

Created by Yoav Luft, 2020. Released under

,

. Created using

, with Jakub Hampl's

for creating the diagram. Many other great packages were used, take a look at elm.json for a complete list.