Other Recommender Research
In addition to the various recommender systems projects I lead, I have also been involved in several side projects with other collaborators.
Surveys and Position Papers
I have written or co-authored a few general recommender system papers, including our survey for Foundations and Trends in HCI:
And position papers on recommender systems research and development, either generally or applied to particular areas:
2016. First Do No Harm: Considering and Minimizing Harm in Recommender Systems Designed for Engendering Health. In Proceedings of the Workshop on Recommender Systems for Health at RecSys '16.and .
2016. Behaviorism is Not Enough: Better Recommendations through Listening to Users. In Proceedings of the Tenth ACM Conference on Recommender Systems (RecSys '16). ACM. DOI:10.1145/2959100.2959179. Acceptance rate: 36% (Past, Present, and Future track).and .
Additional position papers can be found under Reproducible Research.
In this project, led by Tien Nguyen and Daniel Kluver, we examined different interfaces for improving the process of rating movies by giving the user additional information to help guide their rating. We tried several things:
- Showing the user tags related to the movie, to help them recall its characteristics.
- Movies similar to the movie to rate for each of the valid rating values, to provide the user with a point of reference.
- Combining these two interfaces.
The result was published in RecSys 2013.
2013. Rating Support Interfaces to Improve User Experience and Recommender Accuracy. In Proceedings of the Seventh ACM Conference on Recommender Systems (RecSys '13). ACM. DOI:10.1145/2507157.2507188. Acceptance rate: 24%., , , , , , and .
Information Content of Ratings
In this project, led by Daniel Kluver and Tien Nguyen, we attempt to quantify how much information (in the Shannon information theory sense) is contained in a rating of a movie, and use this as the basis for comparing different rating interfaces based on their efficiency (bits per second).
One of the particularly fun developments in this paper is an experimental protocol for estimating a lower bound on the mutual information between ratings and the preference constructs the user's brain, allowing us to reason about the amount of information about preference, not just information, is in a rating. Unfortunately, this protocol requires a ridiculous number of users to achieve any kind of power, but it's a very nice theoretical development in my opinion.
This project, led by Justin Levandoski, embedded recommender technology into an SQL database.
2012. RecStore: An Extensible And Adaptive Framework for Online Recommender Queries Inside the Database Engine. In Proceedings of the 15th International Conference on Extending Database Technology (EDBT '12). ACM, 86–96. DOI:10.1145/2247596.2247608. Acceptance rate: 23%., , , and .