Optimizing "Fairness" - OR Tutorials for Professionals

The purpose of this Tutorials series (cum training) is to cover topics which are not typically covered in academics and are quite relevant for an OR practitioner. Each talk/lecture will be given by a seasoned OR practitioner teaching one topic in-depth. The focus will be on learning practical tips and techniques (keeping the theory to a minimum).

OR Tutorials for Professionals: Session 13

“Fairness” is a difficult concept to define and formulate. It arises in any situation where resources are limited and not all demands can be fully satisfied. In this context, fairness typically means partially unmet demand across recipients in such a way that the outcome is perceived as equitable by all.

For example, Toyota distributors may each request a certain number of vehicles for every model. Similarly, a quick commerce platform such as Blinkit or Swiggy may need to allocate limited inventory across multiple dark stores to meet varying demand. When total demand exceeds supply, the question becomes whether we can create a fair allocation so that no recipient feels unduly disadvantaged.

Modeling fairness is equally challenging. Linear models often fail to capture the desired behavior, while introducing nonlinearity can significantly increase computational run time.

In this talk, we will explore multiple ways to define and model fairness, along with practical tips and tricks. We will also use this example to recap and learn several important mathematical modeling concepts.

Presenter: Arvind Kumar, PhD. (https://www.linkedin.com/in/arviphd/)