Michael Ekstrand

Research Agenda

I study what happens when people and information systems collide. More specifically, my research examines intelligent information systems, such as recommender systems, information retrieval systems and search engines, and user-facing machine learning applications, through the lens of human-computer interaction. Two key questions drive my research agenda:

  1. How can these systems be designed and adapted to better meet human needs in real applications?
  2. How do these systems interact with people, both their individual users and society more broadly?

Recommender systems help people find movies to watch, introduce new friends on social networks, increase online retailers' sales by connecting their customers with personally-relevant products, and direct readers to additional articles on a news publisher’s partner sites. Users interact with recommenders almost everywhere they turn on the modern Internet, and related technologies are impacting their lives in many less visible ways. Their steps through cyberspace and, increasingly, physical space leave data traces that are used in turn to help them, manipulate them, or even incriminate them.

I want to make this world better for the people that it impacts. The aim of my research is to maximize the benefit of recommenders and other intelligent information systems to their users and to society, while identifying and mitigating possible risks that can arise from the use and deployment of these systems.

In pursuit of these goals, I am actively engaged in several ongoing research activities, including:

  • Development and maintenance of LensKit, an open-source recommender systems toolkit with an emphasis on supporting reproducible research in recommender systems.

  • Experiments with public data sets to understand the behavior of recommenders and other artificial intelligence algorithms, particularly with respect to their ability to meet different kinds of user information needs and the emergent properties of their interaction with users.

  • Studies of live user interaction with recommender systems.

Several aspects of this work build on my thesis, Towards Recommender Engineering.

I also maintain interests in programming languages and techniques and computer science education, as well as human-computer interaction and social computing broadly.