Neuro symbolic system detects persuasiveness in online debates
Have You Been Convinced? A Neuro-symbolic Argumentation Reasoning Approach to Detecting Persuasiveness in ChangeMyView Discussions
Computation and LanguageArtificial IntelligenceSymbolic Computation
Summary
People often debate online and try to convince others. This paper shows how a computer system can analyze these debates to figure out if someone's opinion has been changed. The authors combine a smart language model to find arguments with a logical method to see how arguments influence each other. They test the system on a large collection of real discussions and show it works better than some other language models. The system also reveals challenges like dealing with long arguments that reduce its accuracy.
What this means in practice
- •For online community managers: Automatically identify persuasive arguments in discussion forums to understand when user opinions change.
- •For customer support teams: Analyze customer feedback debates to detect influential arguments that could lead to opinion shifts in product perception.
Authors
Siddharth Bhargava, Sara Tonelli, Patricia Martín-Rodilla
Abstract
Argument structures represent how individual arguments interact within a discourse. However, leveraging these structures to infer meaningful knowledge remains challenging, as it requires a formal representation of how argumentative influence propagates across the structure. We propose Neuro-symbolic Argumentation Reasoning System (NARS) for performing persuasiveness detection directly from online discussions. The neural component employs retrieval-augmented iterative LLM reasoning to extract bipolar argument structures, while the symbolic component applies gradual argumentation semantics to propagate argumentative influence over the structure. Using the Webis-CMV-20 corpus, we construct CMV2BAS, a dataset of long, multi-threaded ChangeMyView discussions annotated with LLM-derived bipolar argument structures and labeled with delta events (view change vs.\ no change). We investigate whether gradual argumentation semantics can effectively detect persuasiveness, whether augmenting them with argument quality and graph-based centrality improves performance, and whether symbolic reasoning remains necessary in the presence of modern LLMs. NARS achieves F1=0.57 and AUC=0.56 under its best configuration, outperforming zero-shot Qwen3.6:27B and GPT-oss:20B. We identify support-dominant discussions and long argumentative chains as key failure modes limiting attack influence propagation. Our work demonstrates the feasibility of operationalizing formal gradual argumentation semantics over noisy real-world discourse for scalable macro-level persuasion analysis, while highlighting key challenges and directions for future research in neuro-symbolic argumentation reasoning. We release both CMV2BAS dataset and the NARS framework for public use.