Explaining recommendations in an interactive hybrid social recommender
summary
An interactive hybrid social recommender is built with a Relevance Tuner interface and a version with Relevance Tuner Plus that adds explanation icons showing publication cosine similarities with a ring diagram, topic overlaps via topic modeling and a radar chart, and coauthor paths on a graph. A within subject online study (N=33) finds 90% use the sliders but only 50% click explanations, and explanations raise perceived explainability while lowering perceived controllability and ease of use.
AI-written summary · about
chapters (13)
- 0:00Introduction to hybrid recommenders
- 1:09Problem with static fusion
- 2:13Introducing relevance tuner
- 2:33Adding explanation icons
- 3:15Explaining publication similarity
- 4:21Explaining topic similarity
- 5:30Explaining co-authorship network
- 6:36User study design
- 7:37Information-seeking tasks
- 8:21User feedback on explanations
- 9:26Trade-off in user perceptions
- 10:10Conclusions and future work
- 11:12Q&A and closing
related talks
09mVisual, textual or hybrid: the effect of user expertise on different explanations
16mMaking recommendations explainable
37mInteractive Explainable AI
10mCritiquing for Music Exploration in Conversational Recommender Systems
09mAre Explanations Helpful? A Comparative Study of the Effects of Explanations in AI-Assisted ...
41mIf We Want AI to be Interpretable, We Need to Measure Interpretability
09mWait, But Why?: Assessing Behavior Explanation Strategies for Real-Time Strategy Games
24mAI: Redefining Customer Experience