Atlas · C21 · Social
Empathy & reading emotions
Recognising what others feel and responding in a way that helps.
AI Reliable score
6 / 10
Competent pro level · peak 8 · trend → · grade A
Why this score
On five standard emotional-intelligence tests six LLMs averaged 82% vs 56% for humans. These are text tests: reading a face, voice or room in real time, and felt care, are not measured. Companionship uses are policy-limited.
On the human scale
2Beginner. Recognises explicit emotion words
4Average adult. Average adult: reads feelings in conversation, responds kindly
6Competent pro. Professional: handles upset customers or patients well ← AI today
8Expert. Expert: therapists, skilled nurses
10Best humans. Best human empaths
Future warning
At its best, AI already reaches 8 on this capability. Peak results are never counted in the index; they show where the water may rise next.
Evidence
Schlegel, Sommer & Mortillaro, Communications PsychologyA
E24 · 2025-05
- Measures
- Five standard emotional-intelligence tests
- AI result
- Six LLMs averaged 82% vs humans 56%.
- Humans
- Human participants in the original test validations
- On our scale
- Above average adult on text-based tests; real-time and embodied empathy not measured.
sciencedaily.com ↗
History
First reading: Atlas v0.1, 5 Nov 2026. Every later change will be listed here with its reason.
Disagree? Send a correction
