AI systems, including the most advanced large language models, lack a true understanding of the emotional contexts that humans continuously experience. This gap limits the richness of personalization engines, the predictive power of expert systems and the humanness of chat interactions — a particularly important consideration in sensitive domains like mental health and wellbeing counseling. In sum, without emotional context AI systems fall far short of their full potential.
Coauthored by:
Bill Nolen, PhD, Chief Scientist, AgileBrain
J David Pincus, PhD, Chief Innovation Officer, AgileBrain
Jamie Thompson, CEO, Sprinklenet
AgileBrain addresses this gap with a simple, powerful idea:
Empower AI systems with real-time emotional state data.
Emotions: The Missing Layer in AI
LLMs like OpenAI’s GPT series, Anthropic’s Claude, Meta’s LLaMA, and xAI’s Grok are capable of confident, articulate responses. But their fluency masks a fundamental limitation: they have no real awareness of the human they are interacting with via chatbot or whose behavior they are trying to predict via expert system.
These systems can infer human attributes based on training data, reinforced by human feedback and retrieval of relevant information. But their inferences reflect behaviors learned from large groups, not a specific individual. To date, AI systems have not been able to perceive the actual emotional state of individuals. This blind spot is especially relevant in use cases where emotions predominate, including health and wellbeing, customer experience, consumer marketing and decision support.
Emotional State Data as a First-Class Signal
AI has already transformed industries by ingesting behavioral, environmental, and sensor data to power insights. Why not emotional data? Emotional data can be treated the same way and shows real promise to drive similar transformation.
In fact, multiple large-scale studies have demonstrated the ability of AgileBrain’ metrics to predict emotional states and behaviors with high accuracy, as shown in the diagram below.
