Fairness, Efficiency, and Transparency in ML
Methods
Motivated from the problems faced by under-served communities or in under-resourced settings, we are working to define and quantify fairness in machine learning. Our aim is to build intelligent decision-making systems that effectively trade-off fairness, efficiency, and transparency while being robust to uncertainty/noise in the data.
Sample decision-tree
Applications for Social Good
Design of Optimal and Fair Decision Trees for Decision-Making in Socially Sensitive Settings
We are developing optimal, fair, and interpretable machine learning algorithms suitable for decision-making in socially sensitive settings. Our work is motivated by the increased use of machine learning techniques to make decisions affecting marginalized communities and vulnerable populations.
Census Dataset
Default Dataset
Optimal Classification and Regression Trees
We are also focused on speeding-up computation of optimal decision-trees that can capture arbitrary constraints (e.g., fairnees, interpretability) to faciliate open-world deployment. Our approaches are 50x faster than existing techniques.
Optimality gaps
Number of instances solved
Related Papers
Strong optimal classification trees
(*) S. Aghaei, A. Gomez, P. Vayanos
Under review at Operations Research, January 2021.
Learning optimal and fair decision trees for non-discriminative decision-making
(*) S. Aghaei, (*) M.J. Azizi, P. Vayanos
In Proceedings of 33rd AAAI Conference on Artificial Intelligence, 2019.
note: acceptance rate ~16% in year of submission
Learning optimal classification trees: strong max-flow formulations
(*) S. Aghaei, A. Gomez, P. Vayanos
Technical report, available on Optimization Online, 2020.
(*) M. J. Azizi, P. Vayanos, B. Wilder, E. Rice and M. Tambe
In Proceedings of the 15th International Conference on the Integration of Constraint Programming, Artificial Intelligence, and Operations Research (CPAIOR), 2018.
Invited to Constraints journal fast track for outstanding papers.
Related Grants
CES Triage Tool Redesign and Implementation
Home for Good Foundation
Role: Co-PI (PI: Eric Rice)
Award ID: TBD
Total Award Period Covered: 01/01/2020-12/31/2022
Total Award Amount: $1,450,000
Own Share: $334,000
Designing Fair, Efficient, and Interpretable Policies for Allocating Scarce Resources
USC James H. Zumberge Faculty Research & Innovation Fund Diversity & Inclusion Grant Program
Role: Sole PI
Funded for the period: 07/2018-07/2019
Total award amount: $30,000
Own Share: $30,000