MinorBench: A hand-built benchmark for content-based risks for children

2025-03-13

AI Summary

Existing AI ethics research lacks systematic attention to content risks posed by large language models (LLMs) in child-centric contexts—such as home, school, and peer interactions. Method: Drawing on real-world cases of LLM chatbot misuse by middle-school students, we propose the first content risk taxonomy specifically for minors and introduce MinorBench: the first open-source, manually annotated, interpretable, and multidimensional benchmark designed to evaluate LLMs’ ability to detect and refuse unsafe or inappropriate queries directed at children. Contribution/Results: Empirical evaluation across six mainstream LLMs reveals substantial variation in refusal rates (32%–91%), demonstrating MinorBench’s validity and discriminative power. This work fills a critical gap in child-focused AI safety assessment and provides both empirical grounding and practical tooling for developing child-friendly safety mechanisms.

Technology Category

Machine Learning: Large Multimodal Models (LMMs)Natural Language Processing: Safety and RobustnessPhilosophy and Ethics of AI: Safety, Robustness & Trustworthiness

Application Category

User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendationSearch and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved informationSemantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large Language Models (LLMs) are rapidly entering children's lives - through parent-driven adoption, schools, and peer networks - yet current AI ethics and safety research do not adequately address content-related risks specific to minors. In this paper, we highlight these gaps with a real-world case study of an LLM-based chatbot deployed in a middle school setting, revealing how students used and sometimes misused the system. Building on these findings, we propose a new taxonomy of content-based risks for minors and introduce MinorBench, an open-source benchmark designed to evaluate LLMs on their ability to refuse unsafe or inappropriate queries from children. We evaluate six prominent LLMs under different system prompts, demonstrating substantial variability in their child-safety compliance. Our results inform practical steps for more robust, child-focused safety mechanisms and underscore the urgency of tailoring AI systems to safeguard young users.

Problem

• Address content-related risks for minors using LLMs
• Develop a benchmark to evaluate child-safety compliance in LLMs
• Propose taxonomy and safety mechanisms for child-focused AI systems

Innovation

• Developed MinorBench for child safety evaluation
• Proposed taxonomy for minor content-based risks
• Evaluated six LLMs on child-safety compliance