Check This Out: Recommender Systems

Paper & ToCHI

May 9, 2012 @ 14:30, Room: 19AB

Chair: James Fogarty, University of Washington, USA
AccessRank: Predicting What Users Will Do Next - Note
Community: engineering
Contribution & Benefit: Describes AccessRank, an algorithm that predicts user actions. Log analyses (web visits, window switches, and command use) demonstrate that it outperforms existing techniques (e.g. recency, frequency). Gives directions for deployment.
Abstract » We introduce AccessRank, an algorithm that predicts revisitations and reuse in many contexts, such as file accesses, website visits, window switches, and command lines. AccessRank uses many sources of input to generate its predictions, including recency, frequency, temporal clustering, and time of day. Simulations based on log records of real user interaction across a diverse range of applications show that AccessRank more accurately predicts upcoming accesses than other algorithms. The prediction lists generated by AccessRank are also shown to be more stable than other algorithms that have good predictive capability, which can be important for usability when items are presented in lists as users can rely on their spatial memory for target location. Finally, we present examples of how real world applications might use AccessRank.
ACM
Effects of Behavior Monitoring and Perceived System Benefit in Online Recommender Systems - Note
Community: user experience
Contribution & Benefit: Experiment manipulating an online recommender system's behavior-monitoring functionality and its perceived consumer or corporate benefit. Offers guidance for theorists and designers of recommender systems.
Abstract » Behavior monitoring is an important part of many recommender systems; however, its effects on users' perceptions of such systems are not well understood. We describe a 2x2 factorial experiment that manipulates a simulated recommender system's monitoring of user behavior (monitoring: present vs. absent) and whom the system is perceived to benefit (benefit: corporate vs. consumer). We find that attitudes toward being monitored are moderated by perceptions about system intentions. We propose an explanatory mechanism and highlight the value of understanding the subjective experience of interacting with recommender systems.
ACM
Design and Evaluation of a Command Recommendation System for Software Applications - ToCHI
Contribution & Benefit: Explores the design space of modern recommender systems in complex software applications for aiding command awareness. Performs a 6-week real-time within-application field study in user’s actual working environments.
Abstract » We examine the use of modern recommender system technology to aid command awareness in complex software applications. We first describe our adaptation of traditional recommender system algorithms to meet the unique requirements presented by the domain of software commands. A user study showed that our item-based collaborative filtering algorithm generates 2.1 times as many good suggestions as existing techniques. Motivated by these positive results, we propose a design space framework and its associated algorithms to support both global and contextual recommendations. To evaluate the algorithms, we developed
the CommunityCommands plug-in for AutoCAD. This plug-in enabled us to perform a 6-week user study of real-time, within-application command recommendations in actual working environments. We report and visualize command usage behaviors during the study, and discuss how the recommendations affected users behaviors. In particular, we found that the plug-in successfully exposed users to new commands, as unique commands issued significantly increased.
Asking the Right Person: Supporting Expertise Selection in the Enterprise - Paper
Community: designCommunity: engineering
Contribution & Benefit: Lab study demonstrating that providing additional information about experts in expertise recommenders leads to better selections, and indicating which information is most useful. Offers design implications for expertise recommender creators
Abstract » Expertise selection is the process of choosing an expert from a list of recommended people. This is an important and nuanced step in expertise location that has not received a great deal of attention. Through a lab-based, controlled investigation with 35 enterprise workers, we found that presenting additional information about each recommended person in a search result list led the participants to make quicker and better-informed selections. These results focus attention on a currently understudied aspect of expertise location--expertise selection--that could greatly improve the usefulness of supporting systems. We also asked participants to rate the type of information that might be most useful for expertise selection on a paper prototype containing 36 types of potentially helpful information. We identified sixteen types of this information that may be most useful for various expertise selection tasks.
ACM
To Switch or Not To Switch: Understanding Social Influence in Online Choices - Paper
Contribution & Benefit: Do online recommendations sway people's own opinions? The results of this paper show that this is indeed the case, with important consequences for consumer behavior research and marketing strategies.
Abstract » We designed and ran an experiment to measure social influence in online recommender systems, specifically how often people's choices are changed by others' recommendations when facing different levels of confirmation and conformity pressures. In our experiment participants were first asked to provide their preferences between pairs of items. They were then asked to make second choices about the same pairs with knowledge of others' preferences. Our results show that others people's opinions significantly sway people's own choices. The influence is stronger when people are required to make their second decision sometime later (22.4%) than immediately (14.1%). Moreover, people seem to be most likely to reverse their choices when facing a moderate, as opposed to large, number of opposing opinions. Finally, the time people spend making the first decision significantly predicts whether they will reverse their decisions later on, while demographics such as age and gender do not. These results have implications for consumer behavior research as well as online marketing strategies.
ACM