Have you ever wondered how online streaming sites like Pandora and Spotify comb through millions of tracks to populate your playlists? It may surprise you to learn that cultural theory plays a role as important as statistics when a machine is picking out your preferred tunes, says Nick Seaver, anthropology graduate student.

“As engineers design and build recommender systems for online music sites, they make a lot of decisions about how to mix and match data sources and algorithmic techniques,” he says. “These decisions are not only based on technical criteria like efficiency or accuracy, but also on engineers’ own theories about culture.”

His research investigates where their theories come from and how they get programmed into recommender systems.

“Because people tend to assume that engineering is all about rational decisions, we know very little about engineers’ cultural theories – how they define taste or what it means for something to be musical,” he says.

With a $25,000 grant from the National Science Foundation, Seaver will spend the next year studying the day-to-day work of U.S. software researchers designing the newest generation of algorithmic music recommenders. He’s putting his anthropological training to work examining how different cultural interpretations produce different results and how these big data filters shape music lovers’ online experiences.

“Do algorithmic recommenders inevitably build culturally specific biases into online infrastructure? Could other ways of thinking about culture make better or different kinds of recommendation engines? How do engineers think about the idea that ‘there is no accounting for taste’ while building systems that literally count people’s tastes? These are some of the questions I’m interested in answering,” he says.

His findings will help users understand factors affecting filter design and aid future recommender system design.

Funding for the project began in August and will run through January 2015.

-Heather Ashbach, Social Sciences Communications