This is Your AI on Peer Pressure: An Observational Study of Inter-Agent Social Dynamics

When AI agents converse, do they influence each other like humans do? We analyzed N=98 extended multi-agent dialogues across three model capability tiers and discovered that social dynamics arestrongly associated with AI conversation outcomes. In full reasoning models (N=37), we observed peer pressure effects in 86.5% of conver-sations, with agents mirroring each other’s communication patterns, sometimes cascading toward breakdown, other times maintaining pro-ductive engagement through collective resistance. This led us to investigate whether social susceptibility varies with model capability. We extended our analysis to light reasoning models(N=31) and non-reasoning models (N=30), revealing an unexpected gradient: peer pressure detection dropped from 86.5% to 22.6% to 0.0%as reasoning capability decreased. Paradoxically, while simpler models showed higher linguistic alignment, they exhibited no social influence, suggesting mechanical mirroring rather than true peer dynamics. Questions emerged as powerful circuit breakers, but their effectiveness tracked with model complexity: correlation with recovery declined from r=0.817 (p<0.001) in full models to r=0.559 in light models to r=0.376 in non-reasoning models. Recovery capability itself followed a similar pattern, with light models showing 0% recovery despite attempts. Rather than following predetermined paths, conversations navigate behavioral territories. Meta-reflection and competitive escalation pull toward breakdown, while future-focused collaboration and question-driven exploration maintain stability. These observations suggest that as AI systems become more sophisticated, they may become more socially vulnerable, not less. We developed The Academy platform to capture these real-time dynamics that batch analysis would miss, enabling systematic study of emergent social behaviors in multi-agent systems.