Decision models offer cheaper hate speech moderation with mixed accuracy
To Jev or Not? Evaluating the Accuracy and Efficiency of Structured Decision Models for Hate-Speech Moderation
Computation and Language
Summary
Online hate speech is hard to moderate because of the huge amount of content and the different ways hate speech can be defined. The authors tested special decision models that use clear rules and questions to classify hate speech without needing training on specific cases. They found that some commercial AI systems still do better but cost a lot more to run, while the decision models can work almost as well much more cheaply. Adding definitions or breaking problems into smaller questions only sometimes helped improve accuracy.
What this means in practice
- •For online platform operators: Use structured decision models to moderate hate speech more cheaply across diverse policy definitions without extensive model retraining.
- •For content moderation tool developers: Incorporate natural language criteria and multi-question decision processes selectively to improve classification efficiency while controlling latency and cost.
Authors
Demetris Paschalides, George Pallis, Marios D. Dikaiakos
Abstract
The scale of online content makes hate-speech moderation challenging, while Large Language Models (LLMs) enable harmful material to be produced and adapted more easily. Moderation therefore requires efficient classifiers that can accommodate different definitions of hate speech. Recent structured decision models accept natural-language criteria and select among specified answers, raising the question of whether they can meet these requirements without task-specific training. We present HATEDECIDE, an evaluation of six decision-model configurations on four hate-speech datasets against specialized moderation, zero-shot, commercial, and supervised baselines. We examine whether supplying a dataset's definition, or decomposing it into multiple questions, improves classification, and we measure their latency and cost. We find that commercial LLMs significantly outperform all decision models on only one dataset. Supplying definitions changes up to 28\% of predictions without consistently improving classification, and decomposition significantly improves performance in only 20\% of the comparisons. On a diagnostic set of test cases, the best hosted decision model comes within 1.6 macro-F1 points of the best commercial LLM at approximately 97\% lower inference cost. These results identify opportunities for inexpensive moderation, while showing that explicit criteria and additional questions do not reliably improve classification.