# Understanding Gen Alpha Digital Language: Evaluation of LLM Safety Systems for Content Moderation

**Authors:** Manisha Mehta, Fausto Giunchiglia  
**arXiv:** [2505.10588](https://arxiv.org/abs/2505.10588)  
**Date:** 2025-05-14  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2505.10588v1/

## Key points

- Research by Manisha Mehta and Fausto Giunchiglia evaluates how AI systems interpret Gen Alpha's digital language
- Gen Alpha faces unique online risks due to immersive digital engagement and evolving communication styles
- Study assesses AI models' ability to detect disguised harassment and manipulation within Gen Alpha discourse
- Analysis uncovers critical comprehension failures in existing safety tools for Gen Alpha expressions
- Key contributions include creation of a novel dataset, development of framework for enhancing AI moderation systems, multi-perspective evaluation involving AI systems, human moderators, parents, and input from Gen Alpha co-researchers
- Findings highlight the need to redesign safety systems attuned to youth communication patterns to address vulnerability in online spaces
- Study addresses critical challenges related to digital safety in the context of Generation Alpha's linguistic landscape

## Summary

This research by Manisha Mehta and Fausto Giunchiglia provides a comprehensive evaluation of how artificial intelligence (AI) systems interpret the digital language of Generation Alpha (Gen Alpha), individuals born between 2010 and 2024. As the first cohort raised alongside AI technology, Gen Alpha faces unique online risks due to their immersive digital engagement and evolving communication styles. These may not align with existing safety tools. The study focuses on assessing four prominent AI models - GPT-4, Claude, Gemini, and Llama 3 - in their ability to detect disguised harassment and manipulation within Gen Alpha discourse. The distinct language of Gen Alpha, influenced by gaming, memes, and AI-driven trends, often masks harmful interactions from both human moderators and automated systems. By analyzing a dataset comprising 100 recent expressions from various online platforms such as gaming platforms, social media channels, and video content, the research uncovers critical comprehension failures that have direct implications for online safety measures. Key contributions of this work include the creation of a novel dataset capturing Gen Alpha expressions; the development of a framework aimed at enhancing AI moderation systems for youth protection; a multi-perspective evaluation involving AI systems, human moderators, parents; input from Gen Alpha co-researchers; and an analysis of how linguistic divergence among Gen Alpha individuals increases their vulnerability in online spaces. The findings underscore the pressing need to redesign safety systems that are attuned to youth communication patterns. This is especially crucial given Gen Alpha's reluctance to seek help when adults fail to grasp their digital world. By combining insights from a Gen Alpha researcher with systematic academic analysis, this study addresses critical challenges related to digital safety in the context of Generation Alpha's unique linguistic landscape. The research has been accepted for presentation at ACM FAccT 2025 in Athens and will be published in the conference proceedings before appearing in the ACM Digital Library.