AGI Index dimension · weight 6%
Social intelligence
Modelling other minds and coordinating with them.
Inferring intent, belief and preference from behaviour, and acting well in settings where other agents are doing the same. Includes negotiation, teaching and knowing when to defer.
The argument
Evidence and counter-evidence
Both sides are published at equal weight. A framework that only records what raises a score is not measuring anything.
Raises the score
- Standard theory-of-mind tasks are largely passed, and intent inference from ambiguous instruction has improved.
- Multi-agent coordination succeeds in structured settings with clear protocols.
Holds it down
- Passing theory-of-mind tests written for children is weak evidence for modelling adults with reasons to conceal what they want.
- Calibration is poor: systems rarely know when they have misread a person, and rarely ask.
- Sustained coordination with a partner whose goals shift mid-task remains largely untested.
What the score reads from
Theory-of-mind tasks
Mixed-motive negotiation
Related coverage
Everything bearing on social intelligence
Context
Where this sits against the rest
- Multimodality82
- Coding81
- Tool use76
- Reasoning74
- Memory67
- Generalization66
- Planning61
- Learning59
- Autonomy53
- Scientific discovery48
- Embodiment41
How these scores are producedMethodology v1.2