InterQuest: A Mixed-Initiative Framework for Dynamic User Interest Modeling in Conversational Search
summary
An LLM-based conversational agent named InterQuest dynamically builds a non-hierarchical user model from past interactions, measures confidence via LM self-reflection, and uses a Shannon entropy‑inspired ranking to ask targeted clarifying questions. It presents results in a comparison table and updates user-centric knowledge online. A within-subject study with 18 participants shows InterQuest outperforms two baselines in inferring implicit interests, modeling accuracy, and perceived search quality while resolving three uncertainty types (cold-start, content, scope).
AI-written summary · about
chapters (9)
- 0:00Introduction to InterQuest
- 1:16Problem and motivation
- 3:41Human reasoning insights
- 5:04InnerQuest framework overview
- 6:29Dynamic user knowledge modeling
- 7:33Uncertainty-driven questioning
- 8:15Evaluation and results
- 10:01Design implications and applications
- 10:45Q&A
related talks
21mEvaluating AI Search: A Practical Framework for Augmented AI Systems
03mInterview
13mInteractive Learning and User‑Centered AI: A Retrospective
26mRevolutionizing Search: Unveiling Conversational Search
21mHarnessing Technological Innovation & Accelerating Uptake
32mBeyond Vibe Coding: Leveraging Foundation Models to Infer Intent
21mPersonalize Search Results with OpenSearch Agentic Memory
22mAgentic Hybrid Retrieval: Orchestrating LLM Agents, Hybrid Search & Explainability in OpenSearch