IF-City: Intelligible Fair City Planning to Measure, Explain and Mitigate Inequality

Fairness tools need to be designed for domain experts, not just data scientists: The paper argues that existing AI fairness tools focus too heavily on technical users, but tackling fairness requires involving specialized professionals like urban planners. IF-City was specifically built with urban planners' workflows in mind, suggesting that responsible AI requires domain-specific solutions rather than one-size-fits-all approaches.

AI fairness extends beyond prediction to allocation and planning decisions: While most AI fairness research focuses on predictive decisions (like loan approvals), this work highlights that fair allocation problems—such as where to place parks, schools, and services in cities—require different approaches that integrate multiple constraints and human expertise.

Explainability should answer multiple types of questions to support decision-making: The framework uses three types of explanations—causal (Why did this happen?), contrastive (Why not something else?), and counterfactual (What if we changed this?)—to help planners understand inequality sources and explore solutions, suggesting that responsible AI requires multiple explanation types rather than single answers.

Interactive tools can help identify and mitigate inequality in real urban planning: Testing IF-City with practicing urban planners on an actual New York City neighborhood demonstrated that visual analytics tools can help professionals perceive inequality across different resident groups and generate fairer allocation recommendations through simulation.

Responsible AI for public services requires iterative design with practitioners: The paper emphasizes that fair allocation in city planning cannot be solved algorithmically alone but requires human expertise and iterative refinement, highlighting that responsible AI implementation in public services demands ongoing collaboration between technologists and domain professionals.