Fair Spatial Indexing: A paradigm for Group Spatial Fairness.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Data Science

Background:

• Machine learning (ML) models are increasingly used for critical decisions like loan approvals and hiring.
• Existing fairness research often overlooks geospatial data, despite its potential to introduce bias correlated with protected attributes.
• The widespread use of mobile apps makes location data a significant factor in ML applications.

Purpose of the Study:

• To investigate the impact of location data on fairness in machine learning.
• To propose and evaluate techniques for mitigating location bias and miscalibration in ML models.
• To introduce the concept of spatial group fairness and develop algorithms to address it.

Main Methods:

• Developed a novel spatial indexing algorithm inspired by KD-trees to incorporate fairness considerations.
• Focused on addressing miscalibration issues arising from geospatial attributes in ML.
• Conducted extensive experiments on real-world data to validate the proposed methods.

Main Results:

• The proposed spatial indexing algorithm significantly enhances fairness in ML models.
• The approach effectively mitigates bias introduced by location data.
• High learning accuracy was maintained alongside improved fairness metrics.

Conclusions:

• Location data can introduce significant unfair bias in machine learning systems.
• The developed spatial indexing technique offers an effective solution for achieving spatial group fairness.
• This work highlights the importance of considering geospatial attributes for equitable AI development.