Red-Rec: AI Interface Bridges Feed Repetition and Search
Researchers propose Red-Rec, an AI-supported interface for content exploration when users experience 'vague intent'—feeling their recommendation feed has become repetitive without being able to specify what they want. The system summarizes feed patterns, offers clickable options, asks one follow-up question, and gradually blends new content. A formative study found users recognize staleness but struggle to articulate alternatives, motivating the proactive, low-effort design. Red-Rec was evaluated in a mixed-design lab study against three comparison conditions.
Key facts
- Red-Rec addresses the state of 'vague intent' between passive browsing and active search.
- The system summarizes dominant content categories and latent interests in the current feed.
- It offers clickable exploration options and asks at most one follow-up question.
- New content is gradually blended into the feed after user interaction.
- A formative study showed users recognize feed staleness but cannot articulate alternatives.
- The design is motivated by the need for proactive and low-effort interaction.
- Red-Rec was evaluated in a mixed-design lab study against three comparators.
- The paper is available on arXiv with ID 2605.02902.
Entities
Institutions
- arXiv