Research map / Learning theory and algorithms / Fairness
Promoting Fair Vaccination Strategies through Influence Maximization: A Case Study on COVID-19 Spread
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.
Proposes community-level influence maximization on CBG-POI mobility networks, with fairness variants for race, income, and age risk, to select CBGs for COVID-19 vaccination in three US metro areas.
Problem
Vaccine allocation strategies during COVID-19 that prioritize older people do not address demographic disparities in infections across racial and income groups, and ignore mobility-driven transmission.
Approach
Adapts influence maximization with CELF on a SafeGraph-based CBG-POI mobility network and a stochastic SEIR-style simulator, adding budget-based fairness constraints (equal treatment by race or income), age-risk-weighted influence, and combinations of these to balance infection reduction and fairness.
Key results
- All IM variants outperform RAND and the current-strategy proxy (CS) in reducing infections in every MSA; fairness-constrained variants show no significant loss in infection reduction (e.g., 5% decrease in Philadelphia ~18,000 fewer infections).
- IM-I and IM-IA achieve near-zero KL-divergence for equal treatment across income groups in all MSAs.
- IM-RA (Philadelphia, New York) and IM-IA (Chicago) give the best balance across age, race, and income outcomes.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use influence maximization to select vaccination communities | Vaccination strategy · Philadelphia, New York, and Chicago | infection reduction | Helped |
| Add equal treatment by racial group (IM-R) | IM vaccination · Philadelphia, New York, and Chicago | equal treatment KL-divergence for race | Helped |
| Add equal treatment by income group (IM-I) | IM vaccination · Philadelphia, New York, and Chicago | equal treatment KL-divergence for income | Helped |
| Add age-associated risk weights (IM-A) | IM vaccination · Philadelphia, New York, and Chicago | risk-weighted infection reduction | Helped |
| Combine race equal treatment with age-associated risk weights (IM-RA) | IM vaccination · Philadelphia and New York | infection outcomes across age, race, and income | Helped |
| Combine income equal treatment with age-associated risk weights (IM-IA) | IM vaccination · Chicago | infection outcomes across age, race, and income | Helped |
| Prioritize older communities using age-associated risk weights instead of the current age-prioritization strategy | Vaccination strategy · Philadelphia, New York, and Chicago | risk-weighted infection reduction | Helped |
Surprising findings
- Fairness-constrained IM variants did not significantly reduce overall infection-reduction performance compared to unconstrained IM.
- Achieving equal treatment by income sometimes also yielded fair outcomes by race, and vice versa, suggesting similar fairness objectives in these MSAs.
Limitations
- Results depend on a single mobility-based propagation model calibrated to case counts; greedy submodularity guarantees do not hold for the temporal SEIR model, and the -A variants omit CELF recalculation checks due to submodularity violations.
- Focuses on a single vaccination round and a Western-centric racial framing; does not model vaccine hesitancy.
- Only three US metropolitan areas are evaluated, with improvements in percentage points that are small in absolute terms.
Experimental setup
Datasets: SafeGraph mobility data (Dewey) US Census data (median household income, race) CDC age-group risk factors US COVID-19 case counts for calibration
Benchmarks: Philadelphia, New York, Chicago MSA mobility networks
Metrics: percentage decrease in infections percentage decrease in risk-weighted infections KL-divergence for equal treatment KL-divergence for equal outcome
Baselines: No vaccination Random vaccination (RAND) Current strategy proxy (CS, oldest CBGs first)
Builds on: Kempe et al. influence maximization greedy CELF Chang et al. mobility-based COVID-19 model Farnadi et al. fair influence maximization
Closest papers by meaning
- Equity Promotion in Online Resource Allocation (AAAI 2022 · similarity 0.85)
- Fairness in Social Influence Maximization via Optimal Transport (NeurIPS 2024 · similarity 0.81)
- Bridging Fairness and Efficiency in Conformal Inference: A Surrogate-Assisted Group-Clustered Approach (ICML 2025 · similarity 0.80)
- Controlling The Spread of Epidemics on Networks with Differential Privacy (NeurIPS 2025 · similarity 0.80)
- Learn to Vaccinate: Combining Structure Learning and Effective Vaccination for Epidemic and Outbreak Control (ICML 2025 · similarity 0.79)
- Deploying ADVISER: Impact and Lessons from Using Artificial Intelligence for Child Vaccination Uptake in Nigeria (AAAI 2024 · similarity 0.79)
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.