Persuasion Index: A Theory-Guided Framework for Persuasion Analysis

Persuasion can be systematically measured across contexts: The paper introduces the Persuasion Index (PI), a framework with 15 dimensions grounded in established psychology and communication theories, demonstrating that persuasive techniques follow identifiable patterns that can be detected and analyzed consistently across different types of arguments and domains.

Understanding persuasion helps detect manipulation and improve AI safety: By identifying rhetorical cues used in persuasion, this framework can help detect information manipulation and misinformation, while also supporting the development of safer AI systems that are less vulnerable to or capable of generating deceptive persuasive content.

The framework is transparent and auditable: Unlike "black box" AI systems, PI uses interpretable, rule-based detectors built from lexicons and explicit features (55 sub-features total), making it possible for researchers and practitioners to understand exactly why something is flagged as persuasive rather than trusting opaque algorithmic decisions.

Persuasion techniques vary by topic and context: The analysis reveals that while certain persuasive dimensions show consistent patterns across different datasets, others vary significantly depending on the specific topic and the stance being argued—meaning no one-size-fits-all approach to identifying persuasion exists.

The tool is publicly available for responsible use: The authors released PI as open-source software and a web interface, enabling journalists, researchers, policymakers, and others to analyze persuasive patterns in human and AI-generated communication without requiring specialized technical expertise.