Three Captivating Data Talks from My Week at UNGA

Social Good Summit 2017, on the Topic of Extremism in the Age of Fake News

“You have to reach people when they are sympathetic but not sold” – Yasmin Green, Director of R&D at Jigsaw/Alphabet

Critical to the success of being data driven is the idea of getting the right information to the right people at the right time. At the Social Good Summit Yasmin Green, Director of R&D at Jigsaw, explained Jigsaw’s work around understanding how technology is being used to oppress people or empower individuals. More specifically she went into detail about how her team is using search and video trends to get into the minds of potential ISIS recruits. They learned that to persuade others, simply having more information or knowledge is not enough; instead what is more important is having the right information at the right time. Given that once someone has made their mind it is impossible to convince them otherwise, Jigsaw found that the right time to engage was 6 months prior to the recruit’s decision to join ISIS. If they could reach people, when they are sympathetic but not sold, with more information and data it could help understand all the facts and make better decisions.

Using Data and Technology to Achieve the Sustainable Development Goals High-level Event

During the high-level event at the United Nations Headquarters, Bob Collymore, CEO of Safaricom, stressed the need to understand the ethics around using, sharing and generating data.

During this high-level event at the United Nations Headquarters, Bob Collymore, CEO of Safaricom, stressed the need to understand the ethics around using, sharing and generating data. He particularly mentions social media as this is often a forum where many people willingly give away their data. Privacy abuse is possible, and many times, people are not as well informed as they should be. There are however interesting techniques that I learned later in the week to help combat these issues.

For example, Google Trends data is a real-time dataset proving a unique perspective on what people are searching. I learned that the data is only a sample, on top of being anonymized, categorized and aggregated. Sampling is fine when trying to understand trends, but because you are not seeing 100% of the data the numbers are not exact and this has the potential to create distrust. Another interesting idea came from Bill Howe from the University of Washington on the topic of differential privacy. In general, this means adding noise or synthetic data to a query result to “hide” the contribution of any one individual. In his work, he first derived a model of the real data, added noise and then sampled the noisy model to generate fake data.

Bloomberg Data for Good Exchange 2017, on the Topic of Creating Actionable Data-driven Knowledge in Communities of Need

“The first step of digital literacy is to understand your goal” – Tap Parikh, Professor at Cornell Tech

Throughout the various sessions I attended at the Bloomberg Data for Good Exchange, the theme of data and digital literacy become apparent. On his panel Tap Parikh, Professor at Cornell Tech, made a very critical yet often forgotten remark on the need to first understand the underlying problem or goal you are trying to address before leveraging data to find answers. Especially with the plethora of digital tools and data sources out there, it is critical that we work back from problems to technology and data.

The Data for Democracy side event I attended also touched upon the critical need for education and communication as it relates to data ethics. I really took to the idea of multi-literacy when it comes to data; where a systemic data driven culture does not require everyone to understand coding languages. Similar to computer literacy, you don’t need to know how to code in order to use your laptop or the internet. In fact, you only need to know what is necessary to accomplish your task. What is needed for data literacy is a standard set of definitions and resources, which is critical if we want to extend the data community beyond data scientists. Particularly when it comes to translating data into insights, we need to be able to engage and teach people who don’t know the difference between a correlation and a causation (and the underlying implications). In order to create a more pervasive data-driven society then there needs to a support system that someone can rely on when interpreting and using data.

Danielle Dhillon is the Senior Program Analyst for the Digital Impact Alliance’s Data for Development (D4D) practice, where she works to demonstrate the value of a viable D4D ecosystem for driving effective learning and decision-making across development programs, the public sector and the private sector.