Media Item

News

News || Monday, 30 October 2017

At FHI360, Souktel Shares New Research on Artificial Intelligence for Project Monitoring & Evaluation

The rapid rise of Artificial Intelligence—computers behaving like humans, and performing tasks which people usually carry out–promises to improve everything from travel to personal finance. Can it also transform international development—leading to better project delivery, monitoring, and evaluation?

At MERL Tech, an annual meeting of M&E and tech experts supported by the Rockefeller Foundation & FHI360, Souktel shared its views. In a keynote talk, CEO Jacob Korenblum presented the company’s research in three areas where it believes artificial intelligence (AI) can help enhance M&E:

Object Differentiation

Korenblum first noted that “AI does a good job of telling objects apart”. Souktel has leveraged this capacity to build software systems that track supply delivery more quickly & cheaply: If field staff submit a photo of newly-delivered syringes and another photo of newly-delivered bandages, an AI-based app can automatically check these items off a delivery list—saving huge amounts of time and cost. Still, there are limitations: While AI apps can distinguish between a needle and a BandAid, they can’t yet reliably tell whether the needle is broken, or whether the BandAid is the exact same one that was shipped. These constraints need to be considered carefully.

Natural Language Processing

For anyone who’s sifted through thousands of Excel entries, natural language processing may emerge as the biggest game-changer: Here, Korenblum noted, “AI can interpret large numbers of text responses rapidly, matching them against existing data to find trends—with no need for manual sorting.” Currently, natural language processing works best with simple sentences—so rather than using it to analyze longer, qualitative or open-ended data, Souktel suggests focusing on basic mobile survey response analysis. This alone can reduce data processing time and cost dramatically.

Comparative Facial Recognition

Korenblum observed that this may be the most exciting—and controversial—application of AI: “The potential is huge,” he pointed out: “Qualitative evaluation takes on a whole new meaning when you can capture facial expressions on a phone’s camera.” In this area of AI, Souktel has been focusing on solutions for better attendance tracking: AI is good at determining whether the people in a photo at Time A are the same people in a photo at Time B. Snap a group shot at the end of each community training, and participant tracking becomes significantly easier.

But Korenblum warned that AI applications in this field have a mixed record when it comes to recognizing and matching diverse faces: Each time they scan a new facial image, the applications draw on databases of pre-existing images as a reference point. “Most of those images contain…white men,” as recent MIT research has suggested—so recognition of non-white facial images has been unreliable. This poses a key problem for many regions where aid projects are delivered.

Overall, though, Korenblum noted that Souktel’s work has uncovered several quick wins–areas where AI can measurably improve M&E processes: Simple differentiation between images and objects, tagging and sorting quantitative data (or basic open-ended text), and broad identification of people and groups. And as AI keeps evolving, its ability to make M&E faster, easier, and cheaper will likely grow as well.