Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Beyond Persuasion: Towards Conversational Recommender System with Credible Explanations

Peixin Qin, Chen Huang, Yang Deng, Wenqiang Lei, Tat-Seng Chua

TL;DR

This paper tackles the problem of deceptive or unfaithful explanations in large-language-model–driven conversational recommender systems by introducing PC-CRS, a training-free two-stage framework that ensures explanations are both persuasive and credible. The first stage uses Credibility-aware Persuasive Strategies to guide explanation generation, while the second stage applies iterative self-reflection to remove misinformation. Empirical results on Redial and OpenDialKG show that PC-CRS improves credibility and persuasiveness, and that credible explanations can enhance recommendation accuracy, addressing the trust concerns around LLM-based CRSs. The work highlights the trade-off between persuasiveness and credibility and points to future directions for more personalized strategy selection and broader LLM compatibility to advance trustworthy conversational recommendations.

Abstract

With the aid of large language models, current conversational recommender system (CRS) has gaining strong abilities to persuade users to accept recommended items. While these CRSs are highly persuasive, they can mislead users by incorporating incredible information in their explanations, ultimately damaging the long-term trust between users and the CRS. To address this, we propose a simple yet effective method, called PC-CRS, to enhance the credibility of CRS's explanations during persuasion. It guides the explanation generation through our proposed credibility-aware persuasive strategies and then gradually refines explanations via post-hoc self-reflection. Experimental results demonstrate the efficacy of PC-CRS in promoting persuasive and credible explanations. Further analysis reveals the reason behind current methods producing incredible explanations and the potential of credible explanations to improve recommendation accuracy.

Beyond Persuasion: Towards Conversational Recommender System with Credible Explanations

TL;DR

This paper tackles the problem of deceptive or unfaithful explanations in large-language-model–driven conversational recommender systems by introducing PC-CRS, a training-free two-stage framework that ensures explanations are both persuasive and credible. The first stage uses Credibility-aware Persuasive Strategies to guide explanation generation, while the second stage applies iterative self-reflection to remove misinformation. Empirical results on Redial and OpenDialKG show that PC-CRS improves credibility and persuasiveness, and that credible explanations can enhance recommendation accuracy, addressing the trust concerns around LLM-based CRSs. The work highlights the trade-off between persuasiveness and credibility and points to future directions for more personalized strategy selection and broader LLM compatibility to advance trustworthy conversational recommendations.

Abstract

With the aid of large language models, current conversational recommender system (CRS) has gaining strong abilities to persuade users to accept recommended items. While these CRSs are highly persuasive, they can mislead users by incorporating incredible information in their explanations, ultimately damaging the long-term trust between users and the CRS. To address this, we propose a simple yet effective method, called PC-CRS, to enhance the credibility of CRS's explanations during persuasion. It guides the explanation generation through our proposed credibility-aware persuasive strategies and then gradually refines explanations via post-hoc self-reflection. Experimental results demonstrate the efficacy of PC-CRS in promoting persuasive and credible explanations. Further analysis reveals the reason behind current methods producing incredible explanations and the potential of credible explanations to improve recommendation accuracy.
Paper Structure (22 sections, 4 equations, 8 figures, 11 tables)

This paper contains 22 sections, 4 equations, 8 figures, 11 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. PC-CRS
  4. Strategy-guided Explanation Generation
  5. Credibility-aware Persuasive Strategies
  6. Explanation Generation
  7. Iterative Explanation Refinement
  8. Experiments
  9. Experimental Setup
  10. Main Results
  11. In-depth Analysis
  12. Ablation Study & Human Evaluation
  13. Conclusion
  14. Examples of Credibility-aware Persuasive Strategies
  15. Details on User Simulator and Datasets
  16. ...and 7 more sections

Figures (8)

  • Figure 1: Examples of persuasive and credible explanations. A persuasive and credible CRS would eventually foster the long-term trust to users.
  • Figure 2: Two-stage process of PC-CRS. It first selects an appropriate strategy that is used to generate a candidate explanation. Then, Self-Reflective Refiner eliminates the misinformation in the candidate in an iterative way.
  • Figure 3: Results on relevance gap. It is computed by using metric scores on low credibility explanations to minus high credibility ones. LLM-based CRS caters to user utterances while neglects factual information.
  • Figure 4: Persuasiveness and Credibility scores under different refinement iterations. There is a delicate balance between these two factors.
  • Figure 5: Ablation studies. Both Strategy-guided Explanation Generation (SEG) and Iterative Explanation Refinement (IER) are necessary for PC-CRS.
  • ...and 3 more figures