Fairguard: Harness Logic-based Fairness Rules in Smart Cities

Bias in smart cities operates at granular levels: The research demonstrates that bias in smart city systems persists at "micro-level both temporally and spatially" using real data from Chattanooga, TN, meaning unfair predictions can vary by location and time rather than being uniform across a city.

Two-stage approach to fairness: Fairguard uses both a "static generator" to reduce data bias upfront and a "dynamic component" to monitor and correct predictions in real-time, suggesting that addressing fairness requires intervention at multiple points in the algorithmic pipeline.

Logic-based rules can protect vulnerable groups: The framework uses "temporal logic-based" rules to "guarantee fairness on protected groups at run-time," indicating that explicit, rule-based systems may help ensure that algorithmic decisions don't discriminate against specific populations in smart city applications.

Fairness improvements don't require sacrificing overall performance: The evaluation found that dynamic Fairguard could "guarantee fairness on protected groups at run-time with minimal impact on overall performance," suggesting that cities don't necessarily face a stark trade-off between fair outcomes and system effectiveness.

Smart city bias requires targeted solutions: Rather than applying one-size-fits-all fairness measures, the research indicates that bias in smart cities—which rely on complex sensor networks and temporal-spatial data—needs specialized approaches designed for these specific technical contexts.