Spatiotemporal dynamics of violence in social unrest events based on geo-social media data
Bibliographic record
Abstract
Unrest events occur in the society in various forms. Some of them can escalate into violence, causing severe damage to property, individuals, and society. Recently, social media, together with AI, has become a crucial tool for monitoring and investigating the development and impact of unrest events. This research aims to comprehensively analyze social media data to explore the spatiotemporal patterns of activities and violence associated with unrest events. Highly relevant tweets expressing negative emotions are extracted and analyzed using AI-based natural language processing (NLP) to capture violence within the unrest. Additionally, a data analytics workflow that integrates temporal, spatial, semantic, and network-based methods is employed to provide a comprehensive exploration of the unrest. Using the 2013 Brazil Protests as a case study, we divide the unrest into 5 phases based on the frequency and spatial distribution of negative tweets. During these phases, both the frequency and spatial scope increase, peaking in the 3rd phase before gradually declining. We apply Biterm Topic Modeling (BTM) to identify public concerns during the unrest and analyze their temporal and spatial dynamics. The results reveal that violence is most intense in the 3rd phase and is primarily distributed across southeastern Brazil. Moreover, the spatiotemporal distribution of emotions indicates that fear and anger are the dominant negative emotions when violence occurs, contributing to significant escalations of the unrest. By constructing semantic-temporal networks of negative emotion flows, we pinpoint the leading cities in the unrest. Differences in topic flows within these networks suggest differing motivations behind the unrest, leading to the conclusion that locations where groups expressing concern about violence gather are more likely to become sites of actual violence. These findings can help the government formulate effective measures for managing unrest events and maintaining social stability.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".