Cornell University researchers have developed an algorithm to improve the fairness of online search rankings while retaining their utility and relevance.
Unfairness stems from search algorithms prioritizing more popular items, which means that the higher a choice appears in the list of results, the more likely users are to click on and respond to it, reinforcing one item's popularity while others go unnoticed. When seeking the most relevant items, small variations can cause major exposure disparities, because users commonly select one of the first few listed items.
"We came up with computational tools that let you specify fairness criteria, as well as the algorithm that will provably enforce them," says Cornell's Thorsten Joachims.
The learning-to-rank algorithms, called FairCo, allocate approximately equal exposure to equally relevant choices and avoid preference for items that are already highly ranked. This approach can remedy the unfairness inherent in existing algorithms.
From Cornell Chronicle
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA