Data-driven Analysis of Equity and Fairness in Public Transit
Brief Description of Project:
Micro-transit is increasingly becoming common in urban areas across the globe. Typically, data-driven algorithmic approaches are used to allocates resources to spatial regions of cities. However, existing data is often biased against specific demographics, and using such data reinforces bias in resource allocation. For example, the distribution of rental bikes has been shown to discriminate against features such as socioeconomic status. Crucially, such discrimination is often unintentional and not explicitly modeled in algorithmic approaches. This makes it imperative that existing resource allocation methods from the real world are validated against state-of-the-art standards of fairness, and algorithmic approaches be designed to explicitly take fairness into account. As part of this project, students will work on two major socio-technical problems. First, they would conduct data analysis on open-source transportation data to evaluate fairness of resource allocation. Second, they would evaluate the performance of existing algorithms for resource allocation under constraints that explicitly represent fairness.
The students should have a strong background in data analytics and be comfortable with combinatorial optimization. Knowledge of python is expected.
Nature of Supervision:
Student will work together with the primary investigators and the lab's postdocs. They will be expected to present weekly progress and help in writing the results of the research.
A Brief Research Plan (period is for 10 weeks):
Understanding fairness from the perspective of resource allocation - 1 week.
Understanding the data, features and tools - 2 weeks.
Evaluating fairness in existing data - 3 weeks.
Evaluating approaches under fairness constraints - 3 weeks.
Report writing and Final Presentation - 1 week.
Number of Open Slots: 1
Name: Abhishek Dubey
Department: Electrical Engineering and Computer Science