Synthetic Capstone Spotlight: AidSight
The Master of Information and Data Science (MIDS) program at the UC Berkeley School of Information (I School) culminates with a synthetic capstone project. As part of the I School’s continued commitment to shape the future of data science by promoting its efforts to serve the public good, the I School faculty support MIDS students in their development of capstone projects that allow them to improve human life and benefit society as a whole. One capstone team consisting of MIDS students Nicholas Hamlin, Natarajan Krishnaswami, Glenn Dunmire, and Minhchau Dang came together to create AidSight. AidSight is a platform that uses modern data science techniques to make the 600,000 aid activities reported to the International Aid Transparency Initiative (IATI) understandable and digestible at a glance.
The AidSight team took some time to answer a few questions we had about the impact their platform is having on the global aid and funding community.
Can you give a brief introduction of yourselves?
We met as part of the MIDS program when we joined the January 2014 cohort together. We came from various backgrounds including software development, consulting, and data in the international development world. We quickly formed a team for the capstone project based on our shared interest in working on projects with a focus on social good.
While in the MIDS program, you came up with AidSight for your final project. Can you talk a little about what AidSight is?
When planning development projects, aid professionals typically want to seek out information on similar successful projects. Much like cooking a recipe, it’s valuable for them to understand what has worked for others in the past, what hasn’t, and where they might encounter potential problems. Those funding aid projects also want to understand how their money’s being spent and which organizations it’s flowing through. The IATI registry provides a standard framework that organizations can use to report the details of their work in a scalable way. However, the development world is complex, so it can be hard to fit detail and nuance into a standard like this. The end result is uncertainty among users about whether the data they find is trustworthy enough to use for modeling purposes.
Like other tools, AidSight ingests data published to the IATI registry and helps aid professionals find organizations that have worked on similar projects. What makes AidSight unique is that it clearly visualizes the relationships between organizations and grades the data that those organizations have published to give aid professionals a better sense of how usable that data will be in planning their own projects. This data quality goes beyond just answering the question “Does this dataset comply with the IATI standard?” and instead focuses on “Is this data likely to be useful in practice?”
The I School makes using “data for good” part of its mission. Did that have any impact in shaping the objectives of AidSight?
Regardless of whether they work in tech, finance, medicine, or elsewhere, members of the I School community tend to be keenly focused on how their work can create positive impact beyond just “the bottom line.” Building AidSight in this group of peers helped hone and maintain our focus on delivering concrete practical value to a user group that is typically overlooked or caught up in larger competing interests.
When you came up with this idea, what problem were you looking to solve?
When someone makes a donation to an aid organization, that aid organization can contribute the donations they receive to other organizations, creating a money flow. Initially, we were aiming to bring together multiple data sources in order to get a higher-resolution picture of how money was flowing in the aid sector, and use that picture to answer questions like “If our larger need was to combat terrorism, how are the funding flows addressing that need?”
Before we could hope to tackle that challenge, we had to first get a sense of just how much we could trust all the published data, and we found there was no easy way to do that. Therefore, we pivoted in order to get a better sense of the data we had in front of us, and AidSight was born.
How has AidSight grown in the past few months?
We’re excited about the response that AidSight has received in the aid world. While we haven’t had opportunities to make significant updates to the platform since its launch, we have open sourced all our code so that others can build on the work we’ve done so far. In addition, we’ve had conversations with the IATI team as well as other key members of the open development data community about how they can use what we’ve created to help answer their key questions and incorporate the results into their work. There are a few organizations focused on training others how to leverage the rich world of IATI information, and they’ve added AidSight to their list of recommended ways to interact with the dataset. This led to one of the most exciting pieces of feedback we’ve received since the project concluded — which is that the organizations that are learning about AidSight during these trainings have commented that they feel like it’s the only IATI data platform designed specifically with their needs in mind.
What lies ahead for AidSight? Where do you see it going in the next couple years?
At the moment, we don’t have concrete plans to expand the platform. But, with the interest that we’ve received from the IATI community, we’d love for it to continue to bring value to the aid and philanthropy world whether as a stand-alone app or integrated with other platforms. More broadly, the opportunities for leveraging data science and machine learning techniques in the social sector over the next few years are significant. Solving these problems effectively requires not just technical firepower, but a keen awareness of user needs and the human challenges associated with data use. It’s this balance that the MIDS program is designed to strike, so we’re optimistic about the potential for the future.