Analysing social media posts with artificial intelligence can help experts predict when people will move during crises, allowing for faster and more effective humanitarian aid, according to a new study co-authored by a University of Notre Dame researcher.
The study highlights how powerful computational tools can help address the global displacement crisis, which has seen the number of displaced people nearly double in the last decade. In 2024 alone, one in 67 people fled their homes.
“Traditional data, such as surveys, are extremely difficult to collect during forced migration crises,” said Helge-Johannes Marahrens, assistant professor of computational social science at Notre Dame’s Keough School of Global Affairs. “As early warning systems evolve, artificial intelligence and new digital data can help improve them. Ultimately this can help strengthen humanitarian responses, saving lives and reducing suffering.”
The study, published in EPJ Data Science, analysed almost two million social media posts in three languages on X (formerly Twitter) related to three crises: the 2022 Russian invasion of Ukraine, the 2023 civil war in Sudan, and the economic crises in Venezuela.
Researchers found several key signals:
- Sentiment over emotion: Sentiment (positive, negative, or neutral) was a more reliable signal for predicting movement than emotion (joy, anger, or fear).
- Predicting volume: Sentiment was “particularly helpful” at predicting the timing and volume of cross-border movements.
- AI effectiveness: Pretrained language models, which are trained on massive amounts of text using deep learning, provided the “most effective early warning”.
Marahrens noted the analysis seems to work best in conflict settings like Ukraine, but “not as well” in economic crises that unfold more slowly, such as in Venezuela. He also cautioned that such analyses can trigger “false alarms” and are most valuable as an “early trigger for deeper investigation” combined with traditional data.
Future work could include data from additional social media networks and explore automated translation to analyse more languages.
“Together, these improvements could help strengthen these tools,” Marahrens said, “making them more helpful for policymakers and humanitarian organisations that work with displaced people.”
