We were very honored to welcome representatives from the United Nations Global Pulse organization to our inaugural Datadive in NYC. The UN Global Pulse (UNGP) is a small and highly motivated group within the Executive Office of the Secretary-General that is charged with making sense of the massive amount of data generated and collected by all of the UN's many institutions and directives.
For the NYC Datadive UNGP brought two very interesting data projects. The first was focused on understanding the effects of using fertilizer in Uganda. While we might think that this is a fairly straightforward problem, the interesting element that UNGP brought to the Datadive was a very unique set of data: satellite imagery of plots of land that had been treated with fertilizer and imagery of plots that had not. The project sought to use "ground truth" coordinates for 94 farms in Uganda, half of which applied fertilizer and half of which did not during the first half of 2009, to ask the question: when is the earliest time that we an detect the application of fertilizer?
This was a very interesting project, but also an extremely sophisticated one. This required knowledge of image analysis, as well as classification techniques used for image data. Luckily, several data scientists participating in the Datadive had experience working with satellite data, and a small group worked through the day and night to better understand the problem. Given the scale of the problem, everyone was impressed by how much the group was able to accomplish. To find out more about this project follow the ongoing work at Arizona State University.
The second project focused on the UNGP's Global Well-being Snapshot Mobile Survey. The representatives from UNGP posed the question: If you wanted ask everyone in the
world a question, how would you do it? Their conclusion: mobile phones. The well-being survey asked many questions related to happiness and wellness, as well as socioeconomic questions; such as access to technology. And all of this data was collected via an SMS survey.
For this project, UNGP wanted to explore the data in any ways that the data scientists thought were interesting. There were many categories contained in the data to explore. In the histograms at the right, we see the responses to questions related to "would your pend $15 USD on x..." from a subset of the countries surveyed. For comparison, each respondents are separated by sex. For example, we see that women in Nigeria are much more likely to respond that they would spend this money on entrepreneurship than men, while in Indonesia, men are much more likely to say they would save the money than women. During our presentation the UNGP reviewed many of these observations, which initiated a wonderfully engaging conversation around cultural norms and gender roles in each country.
Along with these summary statistics, the group working with UNGP also produced several maps; each pertaining to a specific survey questions. One advantage of using mobile phones to collect survey data is that each response comes with highly granular geolocation data. From this data the team was able to make detailed maps showing the location of respondents to various survey questions
The two maps featured below are examples of these visualizations. The top map shows the results of the question: how did you feel over the past 7 days? We were particularly heartened by this map, as it showed that in general, people were feeling pretty good all over the world. Of course, there are many questions about the external validity of this data, given that we might think that people in the third-world with access to SMS are of a generally different socioeconomic level than the median household.
On that note, the second map visualizes the responses to the question: in the past 7 days, did you communicate with friends and family using SMS? We can see a clear preference to do so in Southeast Asia, while the responses were generally split in India and South America. Also, we should note that in both of these maps nothing was plotted but the response coordinates themselves. However, from this data a beautiful map of the world emerges.
More recently, the team of data scientists that worked with UN Global Pulse had an opportunity to see their work presented to the General Assembly in a talk about using data for development. The same team that developed the maps above also developed this map, which shows the number of responses to their global well-being snapshot mobile survey.
Working on this data with UNGP was a heartening experience and a great example of what can be done when data is applied for the greater good. The UNGP team brought interesting and enlightening data and problems and it was great to see our data experts working together with theirs to gain greater understanding of what lay beneath the spreadsheets and documents. The results from this weekend were not only useful for informing UNGP about information in their data, but also in informing the international discussion on using data for better understanding of the world at large.
For more detail on the work done with UNGP data, see the project page at our Wiki.