What systems can see
ConversationsPurchasesClicksLocationsPast decisions
The intelligence layer for understanding people
AI has learned the world.Now it must understand a person.
HEBB LABS is building the Human World Model: an evolving understanding of a person’s traits, state, intent and relationships, grounded in context and refined over time.
IdentityStateIntentRelationshipsTrajectory
INDIVIDUAL CONTEXT / t → t+1 HUMAN WORLD MODEL01 / The missing layer
Behaviour is visible. Understanding takes context.
Software can capture what people say, click, buy and choose. It still struggles to distinguish a stable trait from a passing state, an expressed preference from an unspoken need, or a habit from a genuine intention.
What needs context
IntentNeedsStateRelationshipsPossible paths
AI that works with people needs an understanding that can change with them.
02 / The Human World Model
One model. Four connected layers.
Four connected capabilities turn evidence from different sources into a model of the individual, linking what is observed with how a person may change.
Reliable data & evidence
Connects records from sources the user has authorised, preserving their origin and context.
Multimodal causal representation
Separates stable structure, dynamic state and source bias instead of learning surface correlation alone.
Human World Model
Connects long-term traits, current circumstances and possible next events to understand change over time.
Online continual learning
Uses new interactions and later outcomes to update the individual model and inform subsequent assessments.
A model that moves with the person
Living context, not a static profile.
Five dimensions connect enduring traits with changing circumstances, relationships and possible paths through life.
Consented evidence informs a causal representation, which supports an individual temporal Human World Model. Judgements and later outcomes inform continual learning. New evidence returns to refine the same model over time.
01
Identity
Stable traits, values and enduring patterns.
02
State
Current emotion, pressure, readiness and circumstances.
03
Intent
Goals, needs and what may not yet be expressed.
04
Relationships
How people shape one another over time.
05
Trajectory
Events, transitions and plausible next paths.
Research ↔ Reality
Learning from what happens next.
Record an assessment before the outcome. Compare it with what happens next. Use the difference to examine errors, update understanding and identify what still needs to be learned.
01Observe
02Model
03Reason
04Act
05Learn
Reality returns evidence to the model03 / Applications
One foundation. Many applications.
We are building a shared foundation for partners working on products and decisions that depend on understanding people over time.
AI agents & embodied intelligence
Help agents interpret instructions alongside a person’s state and context, so assistance arrives at the right moment.
Personal life companion
Connect context across relationships, career and life decisions, with a model that tracks change and follows up on outcomes.
Leadership transitions
Bring motivation, timing, team relationships and change after a leadership move into decision support.
Character world models
Connect personality, emotional state and story events to keep characters coherent across scenes and interactions.
Social world simulation
Explore how models of individual change can inform population simulations and be tested against observed outcomes.
04 / Compounding advantage
Better evidence. Deeper understanding.
Our advantage comes from connecting rigorous research, evidence gathered over time and a model that can be examined and updated.
Scientific foundation
Published causal methods provide a foundation for testing how and why an inference is made.
Longitudinal evidence
Person, relationship, time and later outcome remain connected.
Continuous understanding
The same individual model becomes more useful as life changes.
Trust
Explicit uncertainty, clear boundaries and user control provide a foundation for lasting trust.
05 / The lab
Research depth. Engineering reach.
HEBB LABS brings highly cited causal AI and world-model research together with reliable learning, agentic systems, open-source engineering and product delivery. The founding team covers all four core capabilities—from reliable evidence to continual learning.
Reliable learningMultimodal causal AIWorld modelsTemporal dynamicsContinual learningAgentic systems
Long-term vision
Human understanding should not remain scarce.
Our aim is to support human judgement with a clearer view of context, change and individual differences. We welcome researchers, investors and organisations who share that ambition.
Begin a conversation