ľ¹ÏÓ°Ôº

Food insecurity continues to threaten the lives of hundreds of millions of people worldwide. According to the United Nations Food and Agriculture Organization, the number of undernourished increased from ~580 million in 2012 to ~730 million in 2022 (). Climate change, population growth, political instability, and economic disparities exacerbate this problem, disproportionately affecting vulnerable communities in developing countries. Predicting these crises is crucial but hindered by incomplete or outdated data sources. By employing natural language processing techniques on news articles, Development Impact group AI pinpoints early signs of food crises, enhancing forecasts by up to 50% 12 months in advance. This provides extra time for proactive measures to be taken. Real-time news integration aids decision-making in data-scarce environments, improving humanitarian responses and saving lives. Future efforts include incorporating local languages to improve predictive accuracy and expand to conflict predictions.

View Publications

 |

View Presentations

Hate speech detection (HSD) models have been created to help address harmful content dissemination online. However, they typically use data from the United States, failing to account for other English dialects worldwide. Conventional testing methods, often involving biased datasets, can also overestimate real-world performance in content moderation. Culturally tailored HSD models combined with human oversight can better identify hateful content. Using Nigerian tweets, Development Impact group AI research found such a model with humans reviewing 1% of tweets flagged as hateful helped moderate 60% of all hateful content. 

Understanding who uses social media can aid in identifying societal trends and improving policy responses. Detection models can address issues such as escalating interethnic violence, aiding policymakers in addressing real-time violence outbreaks. Development Impact group AI uses its best performing HSD model to identify Nigerian users who share hate content on Twitter and target them with ads featuring celebrities sharing prosocial messages. Early findings estimate a 15% drop in hate content shared by those who received the ads compared to a control group, likely because fewer people in their network are sharing hate content as well.

View Publications

 |  |

View Presentations

ImpactAI harnesses the power of large language models (LLMs) to sift through swathes of impact evaluation research and accurately extract insights that provide effortless access to causal evidence for improved decision-making. With an added generative AI layer, this tool lets users provide natural language queries and receive personalized, precise responses. Tailored for a diverse audience committed to global development and humanitarian efforts, primary users include development practitioners, policymakers, international development organizations, academic institutions, research institutes, think tanks, and the economic research community. As a one-stop source that lowers the barriers to accessing high quality research, ImpactAI offers easy access to cutting-edge ideas, best practices, and empirical evidence, thereby enhancing the effectiveness of development projects worldwide. Learn more about this tool and sign-up for access .

View Publications

 | |

View Presentations

ImpactAI: Product Demo