Analysis of social media posts can provide early warning signals of population movement during conflicts and disasters, researchers report. The study, published in EPJ Data Science, tests whether digital data can give signals when surveys and field data are hard to collect.
The team examined nearly 2 million posts in three languages on X and compared methods in three case studies. The cases were Ukraine (10.6 million displaced), Sudan (about 12.8 million displaced) and Venezuela (about 7 million displaced). Researchers found that sentiment labels were a more reliable predictor than emotion labels for timing and volume of cross‑border movements.
Pretrained language models performed best as an early warning tool. Marahrens notes that the method worked better in conflict settings like Ukraine and less well in slower economic crises such as Venezuela. He warns social media analysis can cause false alarms and is most useful as an early trigger alongside traditional data and on‑the‑ground reports.
Difficult words
- analyze — Examine something carefully to understand it.analyzing, analyzed
- insufficient — Not enough; lacking what is needed.
- sentiment — A feeling or opinion expressed.
- predictor — Something that helps to forecast an outcome.
- humanitarian — Concerned with helping people in need.
- displacement — Being moved from one place to another.
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- How can social media improve crisis responses?
- What are the potential risks of using social media data?
- In what ways can data collection be enhanced for aid?
- Why is understanding sentiment important in conflicts?
- How could future research benefit humanitarian efforts?
