![[2602.23123] Multi-Agent Large Language Model Based Emotional Detoxification Through Personalized Intensity Control for Consumer Protection](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.23123] Multi-Agent Large Language Model Based Emotional Detoxification Through Personalized Intensity Control for Consumer Protection
Summary
The paper presents a multi-agent system, MALLET, designed to reduce emotional stimulation from sensational content, enhancing consumer decision-making through personalized intensity control.
Why It Matters
In an era where emotional overload from media can impair rational decision-making, this research offers a novel approach to mitigate such effects. By utilizing advanced AI techniques, it aims to promote healthier information consumption, which is crucial for consumer protection and mental well-being.
Key Takeaways
- MALLET employs four agents to analyze and adjust emotional content.
- The system demonstrated significant reductions in emotional stimulus while preserving semantic meaning.
- Category-level analysis showed varying effectiveness across different news topics.
- Personalized advice based on consumption patterns enhances user experience.
- The framework supports calm information reception without limiting access to original texts.
Computer Science > Artificial Intelligence arXiv:2602.23123 (cs) [Submitted on 26 Feb 2026] Title:Multi-Agent Large Language Model Based Emotional Detoxification Through Personalized Intensity Control for Consumer Protection Authors:Keito Inoshita View a PDF of the paper titled Multi-Agent Large Language Model Based Emotional Detoxification Through Personalized Intensity Control for Consumer Protection, by Keito Inoshita View PDF HTML (experimental) Abstract:In the attention economy, sensational content exposes consumers to excessive emotional stimulation, hindering calm decision-making. This study proposes Multi-Agent LLM-based Emotional deToxification (MALLET), a multi-agent information sanitization system consisting of four agents: Emotion Analysis, Emotion Adjustment, Balance Monitoring, and Personal Guide. The Emotion Analysis Agent quantifies stimulus intensity using a 6-emotion BERT classifier, and the Emotion Adjustment Agent rewrites texts into two presentation modes, BALANCED (neutralized text) and COOL (neutralized text + supplementary text), using an LLM. The Balance Monitoring Agent aggregates weekly information consumption patterns and generates personalized advice, while the Personal Guide Agent recommends a presentation mode according to consumer sensitivity. Experiments on 800 AG News articles demonstrated significant stimulus score reduction (up to 19.3%) and improved emotion balance while maintaining semantic preservation. Near-zero correlation between s...