Research map / Large language models / AI and society
Scalable, Sustainable, Generalizable, and Responsible AI for Public Sector
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.
A talk abstract summarizing the author's AI-for-public-sector work, covering food rescue volunteer engagement with 412 Food Rescue and the Bandit Data-Driven Optimization framework, plus an overview of other applications.
Problem
Public sector organizations face challenges (small data, uniform interventions, incomplete objectives, unintended consequences) that make it hard to deploy AI with measurable societal impact at scale.
Approach
Presents a path toward operationally scalable, financially sustainable, technically generalizable, and socially responsible AI for the public sector, illustrated by a deployed food-rescue notification system and a general prediction-prescription bandit framework.
Key results
- Randomized controlled trial with 412 Food Rescue improved notification hit rate by 44% and rescue claim rate by 9%.
- Bandit Data-Driven Optimization algorithm consistently outperformed a bandit baseline by a margin of 15%-300%.
- Algorithm integrated into 412FR's platform, engaging over 45,000 volunteers and facilitating over 6 million meals per year.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use the online rescue-specific recommender algorithm to send volunteer notifications | 412FR · 412FR volunteer rescue notifications | notification hit rate: 44% | Helped |
| Use the online rescue-specific recommender algorithm to send volunteer notifications | 412FR · 412FR volunteer rescue notifications | rescue claim rate: 9% | Helped |
| Use PRedict-then-Optimize with Optimism in Face of uncertainty | Bandit Data-Driven Optimization · Bandit Data-Driven Optimization | performance: 15%-300% | Helped |
Limitations
- Abstract-level talk outline; methodological details are not provided.
- Results are from specific deployments (412 Food Rescue) and may not generalize without further study.
Experimental setup
Datasets: 412 Food Rescue operational data
Benchmarks: 412 Food Rescue randomized controlled trial
Metrics: notification hit rate rescue claim rate
Baselines: bandit baseline
Builds on: Bandit Data-Driven Optimization PRedict-then-Optimize with Optimism in Face of uncertainty
Closest papers by meaning
- Fostering the Ecosystem of AI for Social Impact Requires Expanding and Strengthening Evaluation Standards (NeurIPS 2025 · similarity 0.83)
- Position: Social Environment Design Should be Further Developed for AI-based Policy-Making (ICML 2024 · similarity 0.79)
- Position: Evaluating Generative AI Systems Is a Social Science Measurement Challenge (ICML 2025 · similarity 0.78)
- No Free Delivery Service: Epistemic limits of passive data collection in complex social systems (NeurIPS 2024 · similarity 0.78)
- Position: Technical Research and Talent is Needed for Effective AI Governance (ICML 2024 · similarity 0.78)
- Position: The Right to AI (ICML 2025 · similarity 0.78)
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.