MASCOT: Towards Multi-Agent Socio-Collaborative Companion Systems
This work addresses the challenge of maintaining diverse and productive interactions in multi-agent systems for emotional and cognitive support, offering a practical solution for building socially intelligent companions.
The paper tackles the problems of persona collapse and social sycophancy in multi-agent socio-collaborative companion systems by proposing MASCOT, a framework that uses bi-level optimization to align individual personas and enhance group dialogue, resulting in improvements of up to +14.1 in Persona Consistency and +10.6 in Social Contribution over state-of-the-art baselines.
Multi-agent systems (MAS) have recently emerged as promising socio-collaborative companions for emotional and cognitive support. However, these systems frequently suffer from persona collapse--where agents revert to generic, homogenized assistant behaviors--and social sycophancy, which produces redundant, non-constructive dialogue. We propose MASCOT, a generalizable framework for multi-perspective socio-collaborative companions. MASCOT introduces a novel bi-level optimization strategy to harmonize individual and collective behaviors: 1) Persona-Aware Behavioral Alignment, an RLAIF-driven pipeline that finetunes individual agents for strict persona fidelity to prevent identity loss; and 2) Collaborative Dialogue Optimization, a meta-policy guided by group-level rewards to ensure diverse and productive discourse. Extensive evaluations across psychological support and workplace domains demonstrate that MASCOT significantly outperforms state-of-the-art baselines, achieving improvements of up to +14.1 in Persona Consistency and +10.6 in Social Contribution. Our framework provides a practical roadmap for engineering the next generation of socially intelligent multi-agent systems.