Aligned Agents, Biased Swarm: Measuring Bias Amplification in Multi-Agent Systems

The project addresses a critical and emerging gap in AI safety: the transition from individual agent alignment to collective 'swarm' alignment. While individual LLMs are heavily tuned (RLHF), the project demonstrates that the structure of their interaction (topology) can cause bias to re-emerge or amplify. Despite the theoretical importance, the project currently has a defensibility score of 2 because it is a very new (4 days old) research-oriented repository with zero stars and no community adoption yet. Its value lies in the methodology and the specific empirical findings rather than a software moat. Frontier labs like OpenAI (with their 'Swarm' framework) and Microsoft (with AutoGen) are moving rapidly into MAS; while they care about safety, they are more likely to develop proprietary telemetry for bias than adopt an external academic framework. The primary risk is that this research becomes a standard citation in the field but the code itself remains a static reference implementation rather than a tool used in production pipelines.