Paper Number

ICIS2026-1958

Paper Type

Completed

Abstract

Conversational recommender systems (CRSs) struggle with vague user intent: existing LLM-based approaches rely on reactive questioning, underutilize experience from historical dialogues, and lack decision-theoretic sequential optimization. Adopting a design science paradigm, we develop the Socratic Collaborative Conversational Recommender System (SCCRS), a multi-agent artifact that reconceptualizes recommendation as preference co-construction under partial observability. SCCRS integrates (1) Socratic preference elicitation that actively scaffolds user sensemaking to clarify vague user intents, (2) collaborative experience retrieval that transforms historical trajectories into pseudo-gradient signals guiding dialogue planning, and (3) a POMDP formalization balancing exploration and exploitation. LLM-simulated evaluations on REDIAL and OpenDialKG show that SCCRS significantly outperforms state-of-the-art baselines in recommendation success, dialogue efficiency, and ranking accuracy. This study contributes a systematically evaluated artifact and design principles for adaptive, experience-driven conversational decision support.

Comments

10_Computational_design

Share

COinS
 
Dec 14th, 12:00 AM

Towards Collaborative Conversational Recommendation: A Multi-Agent Framework with Socratic Intent Elicitation and Experience Retrieval

Conversational recommender systems (CRSs) struggle with vague user intent: existing LLM-based approaches rely on reactive questioning, underutilize experience from historical dialogues, and lack decision-theoretic sequential optimization. Adopting a design science paradigm, we develop the Socratic Collaborative Conversational Recommender System (SCCRS), a multi-agent artifact that reconceptualizes recommendation as preference co-construction under partial observability. SCCRS integrates (1) Socratic preference elicitation that actively scaffolds user sensemaking to clarify vague user intents, (2) collaborative experience retrieval that transforms historical trajectories into pseudo-gradient signals guiding dialogue planning, and (3) a POMDP formalization balancing exploration and exploitation. LLM-simulated evaluations on REDIAL and OpenDialKG show that SCCRS significantly outperforms state-of-the-art baselines in recommendation success, dialogue efficiency, and ranking accuracy. This study contributes a systematically evaluated artifact and design principles for adaptive, experience-driven conversational decision support.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.