SOCIAL INTELLIGENCE FOR ROBOTS

Robots know what to do.
We help them understand us.

Social Minds is a hardware-independent social cognition layer for robots. It interprets social situations, maintains a relational model, and helps existing controllers act with awareness of people, context and norms.

person
person
person
robot
ready?
waiting
handover
social
context

THE GAP

Capability is not the same as social competence.

Robots can navigate, manipulate, plan and increasingly reason. But environments designed for people demand more than task completion. Timing, attention, distance, intention, expectations and relationships all change what the appropriate action is.

01

Read the room.

Infer readiness, attention, intent, affect and the interaction already unfolding around the robot.

02

Understand relationships.

Maintain a relational view of people, roles, histories, expectations and what matters between them.

03

Act appropriately.

Modulate timing, approach, distance, handovers and recovery without replacing the robot's existing abilities.

THE SOCIAL LAYER

One social layer.
Any robot. Any task.

Social Minds sits beside a robot's existing controller. It observes what is socially relevant, maintains a relational world model, runs social inference, and turns that understanding into constraints, modulation and new behaviours.

01 Perceive people · gaze · pose · events
02 Relational world model entities · relations · context · history
03 Infer intent · affect · norms · scripts
04 Shape behaviour modulate · constrain · evoke
Built to complement
VLAs planners ROS systems classical controllers custom stacks

TECHNOLOGY

Built from social cognition,
not just prompts.

Social behaviour is not a single prediction. It emerges from multiple processes operating at different timescales over the same relational world.

Our architecture draws on active inference, theory of mind, affect, social norms, interaction scripts and relational world modelling. Different inference processes can operate asynchronously while sharing a common account of who is present, what is happening and what it means socially.

WHERE IT MATTERS

Robots are entering human spaces.
They need to understand the humans in them.

01

Hospitality & service

Know when to approach, when to wait, how to offer, and how to recover gracefully after interruption.

02

Collaborative work

Coordinate with people as partners rather than treating them as moving obstacles in a task environment.

03

Care & assistance

Use social context, attention and relationships to support more predictable, legible and respectful interaction.

RESEARCH → DEPLOYMENT

Social intelligence has to be engineered.

Social Minds AI is being developed from research in social robotics, active inference, cognitive science and artificial intelligence, with roots in Sheffield and collaborations across disciplines.

We are turning that research into a practical, hardware-independent layer that can be tested on real robots in real social environments.

BUILD WITH US

Making robots social.

We are looking for robotics partners, pilots and collaborators who want to test social intelligence in real environments.

Contact Social Minds