Nick Seaver, Tufts University
Overwhelmed, Alone, and Uncool: Visions of the Listener in Algorithmic Recommendation
Who are recommender systems for? Over the relatively short history of algorithmic recommendation, the answer to this question has evolved. At their academic origins in the mid-1990s, researchers working on the new “collaborative filters” envisioned them as aids to eager but overloaded users—music enthusiasts bewildered by the choices available to them or business people overwhelmed by email. Today, recommender systems are a ubiquitous feature of the commercial web, filtering social media posts, personalizing the delivery of news, suggesting movies, therapists, restaurants, and countless other choices to users. Now, the prototypical imagined user is not an enthusiast, but an indifferent person, who might be tempted into using a company’s services by well-crafted personalization. This talk draws on ethnographic research with the developers of music recommender systems in the US to describe their vernacular social theory—the understanding of users that informs the choices they make in designing and deploying software systems that, in recent years, have been ascribed great political and social power.