By Peter Berton

IOWA CITY, Iowa – Dating sites put users through all sorts of hoops in an effort to match them with others who may be compatible. Sometimes the digital machinations work … and sometimes they don’t.

University of Iowa researcher Kang Zhao and his team are trying to improve the odds. They have created a “collaborative filtering” model that assesses not only a user’s preferences when picking others, but also a nebulous characteristic that might best be described as “interpersonal magnetism” — his or her success at being picked by others. The engine then uses the data to recommend people who are the most likely to respond favorably to a given member’s contact requests.

According to Zhao’s description, the filter works something like Amazon’s recommendation engine, which suggests additional merchandise based on past purchases. Zhao’s engine attempts to recommend new contacts based on users’ experiences with previous contacts and those contacts’ experiences with their own contacts, ad infinitum. Characteristics possessed in common by successful contacts go into a “plus” column in the software’s data engine, which searches for other members with similar profiles.

Essentially, the software seems to believe, if a user’s previous successful contacts liked beets, the date-seeker will be more successful if he seeks out more beet-lovers.

Zhao claims the system has performed well in tests that measured both the software’s ability to suggest appropriate matches and reciprocity rate, or the number of matches who responded positively to a user’s initial contact. The research report, “User Recommendation in Reciprocal and Bipartite Social Networks – A Case Study Of Online Dating,” may be accessed here [PDF].