Contextualized Counterspeech: Strategies for Adaptation, Personalization, and Evaluation

2024-12-10

arXiv.org

AI Summary

Current AI-generated counterspeech lacks contextual adaptability to moderation scenarios and user characteristics, hindering precise intervention. To address this, we propose the first contextualized and personalized counterspeech generation paradigm. Our approach builds upon LLaMA2-13B and integrates multi-source contextual signals—including violation type, user history, and platform-specific policies—into a unified situational awareness model. We conduct a pre-registered hybrid crowdsourcing experiment for human-AI collaborative evaluation. Results demonstrate that our method significantly outperforms generic baselines in both informational sufficiency and persuasive efficacy. Human annotator agreement reaches substantial inter-rater reliability (Cohen’s κ = 0.72), while conventional automated metrics exhibit significant misalignment with subjective effectiveness—revealing critical limitations of existing evaluation frameworks and motivating a new human-in-the-loop assessment standard.

Technology Category

• Humans and AI: Intelligent User Interfaces
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Generation

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

AI-generated counterspeech offers a promising and scalable strategy to curb online toxicity through direct replies that promote civil discourse. However, current counterspeech is one-size-fits-all, lacking adaptation to the moderation context and the users involved. We propose and evaluate multiple strategies for generating tailored counterspeech that is adapted to the moderation context and personalized for the moderated user. We instruct an LLaMA2-13B model to generate counterspeech, experimenting with various configurations based on different contextual information and fine-tuning strategies. We identify the configurations that generate persuasive counterspeech through a combination of quantitative indicators and human evaluations collected via a pre-registered mixed-design crowdsourcing experiment. Results show that contextualized counterspeech can significantly outperform state-of-the-art generic counterspeech in adequacy and persuasiveness, without compromising other characteristics. Our findings also reveal a poor correlation between quantitative indicators and human evaluations, suggesting that these methods assess different aspects and highlighting the need for nuanced evaluation methodologies. The effectiveness of contextualized AI-generated counterspeech and the divergence between human and algorithmic evaluations underscore the importance of increased human-AI collaboration in content moderation.

Problem

• AI counterspeech lacks context and personalization
• Tailored counterspeech improves adequacy and persuasiveness
• Human-AI collaboration enhances content moderation evaluation

Innovation

• LLaMA2-13B model fine-tuning
• contextualized counterspeech generation
• mixed-design crowdsourcing evaluation