Causal discovery and counterfactual reasoning to optimize persuasive dialogue policies
Bibliographic Data
| ID | 21452227 |
|---|---|
| Authors | Donghuo Zeng (0000-0002-6425-6270, KDDI Research, Inc., corresponding author), Roberto Legaspi (0000-0001-8909-635X, KDDI Research, Inc.), Yuewen Sun (0009-0002-4842-1608, Carnegie Mellon University), Xinshuai Dong (Carnegie Mellon University), Kazushi Ikeda (0000-0003-3330-6121, KDDI Research, Inc.), Peter Spirtes (0000-0002-1385-190X, Carnegie Mellon University), Kun Zhang (0000-0003-1636-5373, Carnegie Mellon University) |
| Year | 2026 |
| Volume | 45 |
| Issue | 8 |
| Pages | 1515-1529 |
| Publication date | 2026-05-09 |
| Peer Reviewed | Yes |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | Behaviour and Information Technology (JOURNAL) |
| Journal identifiers | ISSN: 0144-929X • E-ISSN: 1362-3001 |
| Publisher | Informa UK Limited (PUBLISHER • GB) |
| DOI | 10.1080/0144929x.2025.2478276 |
| OpenAlex | W4408648820 |
| Language | EN |
| References cited | 28 |
Tailoring persuasive conversations to users leads to more effective persuasion. However, existing dialogue systems often struggle to adapt to dynamically evolving user states. This paper presents a novel method that leverages causal discovery and counterfactual reasoning for optimising system persuasion capability and outcomes. We employ the Greedy Relaxation of the Sparsest Permutation (GRaSP) algorithm to identify causal relationships between user and system utterance strategies, treating user strategies as states and system strategies as actions. GRaSP identifies user strategies as causal factors influencing system responses, which inform a Bidirectional Conditional Generative Adversarial Network (BiCoGAN) in generating counterfactual utterances for the system. Subsequently, we use the Dueling Double Deep Q-Network (D3QN) model to utilise counterfactual data to determine the best policy for selecting system utterances. Our experiments with the PersuasionForGood dataset show measurable improvements in persuasion outcomes using our approach compared to baseline methods. The observed increase in cumulative rewards and Q-values highlights the effectiveness of causal discovery in enhancing counterfactual reasoning and optimising reinforcement learning policies for online dialogue systems
Counterfactual conditional · Counterfactual thinking · Economics · Positive economics · Advanced Software Engineering Methodologies · Advanced Text Analysis Techniques · Computer Science · Multi-Agent Systems and Negotiation · Psychology · Social Psychology
| Citation velocity | historical |
|---|---|
| Highly cited | No |