AI agents need more than memory.
Towards social intelligence for AI agents.
We’re exploring how AI agents can reason about relationships, navigate different perspectives, and anticipate the social consequences of their actions over time.
Intelligence is inherently social.
People see the same situation differently. Relationships evolve. Trust depends on context. Expectations and commitments outlast individual conversations.
As AI agents take on more complex roles alongside people, organisations and other agents, they need to make sense of this social world — and how their actions change it.
Towards social intelligence.
- Social context
- How can agents keep track of relationships, roles, beliefs and expectations as interactions unfold?
- Social cognition
- How can they reason about different perspectives, intentions, social norms and uncertainty?
- Social behaviour
- What helps an agent honour commitments, respond to change and remain coherent across long-running interactions?
A new layer for agent systems.
We’re developing the foundations for social intelligence infrastructure that works alongside existing foundation models and memory systems.
Our aim is to give developers the tools to build agents capable of understanding and navigating complex social environments.
Let’s explore this together.
We’d like to hear from people working on social intelligence, agent architectures, or collaborative AI.