Cornell researchers created a fairer system for recommendations—from hotels to jobs to videos—so a few top hits don’t get all the attention.
The new ranking system still offers relevant options, but more fairly distributes users’ attention to the search results. It can be applied to online markets such as travel sites, recruitment platforms and news aggregators.
Yuta Saito, PhD student in computer science, and Thorsten Joachims, professor of computer and information science at Cornell Ann S. Bowers College of Computing and Information Science, describe their new system in “Fair Ranking as Fair Division: Impact-Based Individual Fairness in Ranking” , published in the Proceedings of the 2022 Association for Computing Machinery Special Interest Group on Knowledge Discovery and Data Mining Conference.
“Anyone who ranks highly in recommendation systems and search engines benefits greatly from this,” says Joachims. “The user’s attention is a limited resource and we need to allocate it fairly to the items.”
Traditional recommender systems try to rank articles purely based on what users want to see, but many articles get unfairly low places in the order. Items of similar value can vary widely in rankings, and for some items, the odds of being spotted on a platform are worse than random.
To solve this problem, Saito developed an improved ranking system based on ideas from economics. He applied principles of “fair sharing” – how to share a limited resource, such as food, fairly among the members of a group.
Saito and Joachims demonstrated the feasibility of the ranking system using synthetic and real data. They found that it provides the user with viable search results while meeting three fair splitting criteria: each item’s advantage of being ranked on the platform is better than being discovered by accident; the effect of an element such as B. Revenue cannot be easily improved; and no item would gain an advantage by changing its ranking relative to other items in a series of searches.
“We completely redefined fairness in ranking,” said Saito. “It can be applied to any type of two-sided ranking system.”
For example, if deployed on YouTube, the recommendation system would present a more diverse stream of videos and potentially distribute revenue more evenly among content creators. “We obviously want to please the users of the platform, but we should also be fair to the video creators in order to preserve their diversity in the long term,” Saito said.
For online recruitment platforms, the fairer system would diversify search results instead of showing all employers the same top candidates.
In addition, the researchers suggest that this type of recommendation system could also help viewers discover new films to watch online, allow scientists to find relevant presentations at conferences, and provide consumers with a more balanced choice of news stories could.
The National Science Foundation and the Funai Foundation for Information Technology funded the research.
Patricia Waldron is a writer at Cornell Ann S. Bowers College of Computing and Information Science.