Explore dynamic pricing strategies with fairness constraints in this 52-minute coffee talk from GERAD Research Center. Delve into the challenges of personalized pricing while addressing concerns about fairness from customers and policymakers. Learn about two types of fairness constraints: price fairness and demand fairness. Discover an algorithm based on an infrequently-changed upper-confidence-bound (UCB) method for achieving near-optimal regret performance. Examine the extension to non-stationary demand and its relevance to preventing price gouging. Investigate demand fairness requirements and a primal-dual learning framework algorithm designed to achieve similar demand levels across different customer groups. Gain insights into personalized pricing models, fairness legislation, and the structure of pricing algorithms with fairness constraints.
Overview
Syllabus
Intro
A personalized pricing model
Application of personalized pricing
Legislation to guarantee fairness
This research focuses on pricing algorithms
Fairness constraints in this research
Structure of today's talk
Model of personalized pricing
Formal definition of price fairness
Benchmark and regret
Challenges from group fairness: classical algorithms no longer work
Challenges from time fairness: cannot update price frequently
Algorithm: Fair Pricing with UCB (FaPU)
Performance of FaPU
Price gouging when demand is non-stationary
Model non-stationary demand
Idea of Algorithm FaPU-Master
Regret of Algorithm FaPU-Master
Brief Discussion of Demand Fairness
Taught by
GERAD Research Center
