LLM-based Rewriting of Inappropriate Argumentation using Reinforcement Learning from Machine Feedback
This addresses the challenge of civil online discussions for social media platforms by preventing negative behavior during content creation, offering a more efficient alternative to post-hoc moderation.
The paper tackles the problem of inappropriate language in online arguments by proposing a reinforcement learning-based rewriting approach that uses an LLM as an initial policy, which significantly outperforms baselines including few-shot learning, prompting, and humans in mitigating inappropriateness while preserving content.
Ensuring that online discussions are civil and productive is a major challenge for social media platforms. Such platforms usually rely both on users and on automated detection tools to flag inappropriate arguments of other users, which moderators then review. However, this kind of post-hoc moderation is expensive and time-consuming, and moderators are often overwhelmed by the amount and severity of flagged content. Instead, a promising alternative is to prevent negative behavior during content creation. This paper studies how inappropriate language in arguments can be computationally mitigated. We propose a reinforcement learning-based rewriting approach that balances content preservation and appropriateness based on existing classifiers, prompting an instruction-finetuned large language model (LLM) as our initial policy. Unlike related style transfer tasks, rewriting inappropriate arguments allows deleting and adding content permanently. It is therefore tackled on document level rather than sentence level. We evaluate different weighting schemes for the reward function in both absolute and relative human assessment studies. Systematic experiments on non-parallel data provide evidence that our approach can mitigate the inappropriateness of arguments while largely preserving their content. It significantly outperforms competitive baselines, including few-shot learning, prompting, and humans.