Research map / Label-efficient and robust learning / Recognition and long-tail
Novax or Novak? Estimating Social Media Stance towards Celebrity Vaccine Hesitancy (Student Abstract)
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Builds a BERT-based stance classifier on actively labeled YouTube comments to measure public support for and opposition to Novak Djokovic's COVID-19 vaccine stance after his 2022 Australia deportation, and tracks lingering support over time.
Problem
Measuring public stance toward a celebrity's vaccine hesitancy on social media, including the spike around Djokovic's deportation and whether support lingers as he keeps winning tennis titles.
Approach
Applies active learning (certainty sampling across YouTube datasets, with GPT-4 and two human annotators for labeling) to train a stance classifier that tracks celebrity vaccine stance discourse over time and across channel types (tennis, news, vaccine skeptic).
Key results
- The BERT stance classifier achieves a Macro F1 of 78.0 ± 0.4.
- Supporters of Djokovic's vaccine stance outnumbered opponents by over 4x around the deportation episode.
- Tennis channels have largely moved on, while news networks and vaccine-skeptic opinion videos still show many comments supporting Djokovic's stance.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Compare supportive and opposing comments around Djokovic's deportation | YouTube discourse about Djokovic's vaccine stance · Dvaccine, July 1, 2021 to July 1, 2022 | stance distribution: over 4x | Mixed |
| Compare discussion across tennis channels, news networks, and vaccine-skeptic videos | Djokovic vaccine-stance discussion · Dtennis | supportive stance prevalence | Mixed |
Surprising findings
- Public support for Djokovic's vaccine stance outweighed opposition, partly attributed to the Australian government's handling of his immigration.
Limitations
- Relies on only two human annotators plus GPT-4 for labeling, with limited annotation quality analysis in the abstract.
- Focuses on a single celebrity and a single platform (YouTube).
- Stance classifier performance (Macro F1 78.0) leaves room for error in downstream trend estimates.
Experimental setup
Datasets: Dvaccine Dtennis Dstance
Benchmarks: Dstance (held-out annotated YouTube comments)
Metrics: Macro F1
Builds on: active learning certainty sampling BERT
Closest papers by meaning
- Promoting Fair Vaccination Strategies through Influence Maximization: A Case Study on COVID-19 Spread (AAAI 2024 · similarity 0.73)
- Bridging Fairness and Efficiency in Conformal Inference: A Surrogate-Assisted Group-Clustered Approach (ICML 2025 · similarity 0.71)
- Deploying ADVISER: Impact and Lessons from Using Artificial Intelligence for Child Vaccination Uptake in Nigeria (AAAI 2024 · similarity 0.71)
- Characterizing Information Seeking Events in Health-Related Social Discourse (AAAI 2024 · similarity 0.71)
- Learn to Vaccinate: Combining Structure Learning and Effective Vaccination for Epidemic and Outbreak Control (ICML 2025 · similarity 0.70)
- AutoCast++: Enhancing World Event Prediction with Zero-shot Ranking-based Context Retrieval (ICLR 2024 · similarity 0.70)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.