I’ve spent forty years building technology. AI architectures. Computing systems. Products that millions of people use. But lately, I’ve been haunted by a simple question that none of my technical expertise can answer:
What does a machine actually understand when it claims to recognize emotion?
The question started innocently enough. I’d see headlines about “emotionally intelligent AI” and “empathetic chatbots.” Conference speakers would confidently declare that AI was learning to “read our feelings.” Investors were pouring money into emotional recognition startups.
And yet, when I looked beneath the surface, I saw something troubling.
The Moment I Realized We’re Fooling Ourselves
It hit me during a demonstration of an “emotion detection” system. The AI confidently labeled a CEO’s expression as “confident” during a quarterly earnings call. Everyone in the room nodded approvingly.
But I’d analyzed that same CEO extensively for my writing. I knew his patterns. That wasn’t confidence – it was controlled panic. The slight tension around his eyes. The forced steadiness in his voice. The strategic pauses that bought him time to think.
The AI saw “confidence” because that’s what the training data labeled as confidence. It was reading performance, not truth.
That’s when it clicked: We’re not building emotional intelligence. We’re building emotional imitation.
And almost nobody is talking about the difference.
The Beautiful Lie We Keep Telling
The narrative around Emotional AI is seductive. We’re promised machines that will understand us, connect with us, maybe even help us navigate our own feelings.
But here’s what actually happens:
A model predicts “sadness” because someone once labeled similar pixels as sad. A chatbot sounds empathetic because its training data included empathetic-sounding text. A system detects stress because the dataset said those biometric patterns meant stress.
We’re not modeling emotion. We’re reproducing labeled patterns.
These are fundamentally different things.
And the gap between them is enormous – and growing.
What We’re Missing About Human Emotion
Real emotion isn’t a label or a classification. It’s contextual, relational, contradictory, and deeply personal.
When a leader shows anger in a boardroom, is it actual anger? Strategic positioning? Defensive fear? All three at once?
When someone smiles during a difficult conversation, are they happy? Uncomfortable? Performing politeness? Hiding something else entirely?
Emotions are not signals. They’re stories.
They’re shaped by power dynamics, past experiences, cultural norms, relationship histories, and a thousand invisible factors that no dataset can capture.
Yet we’re building AI systems that flatten this complexity into probability scores and output classes.
We have tools that can classify faces. We have chatbots that can mimic empathy. We have models that score sentiment.
But we still can’t answer the human questions that actually matter:
What is someone really feeling? Why? What are they hiding? How does their past shape their present reaction?
The models don’t know. And they won’t magically know just because we make them bigger.
Why This Matters More Than We Think
This isn’t just philosophical hand-wringing. There are real consequences.
When leaders use emotional-AI tools thinking they’re measuring emotional truth, they’re actually measuring emotional performance. They’re rewarding people who look emotionally appropriate, not people who are emotionally genuine.
They’re reinforcing stereotypes. Punishing culturally diverse expression. Misreading vulnerability. Overvaluing polish.
And because the AI outputs look objective – numbers, percentages, confidence scores – people trust them.
This is the dangerous illusion we must break.
What I’m Building Instead
For months, I’ve been writing deep emotional analyses of leaders and organizations. Readers tell me they find something there that typical business analysis misses – emotion as a structural force, not a soft variable.
But public writing has limits. It must be polished, coherent, finished. It can’t show the messy middle. The half-formed ideas. The experimental failures.
So I’m building something different: EQ Engine.
Not a product. Not a startup. Not another emotional-AI tool making bold claims.
A lab. A living research environment.
A place where:
- We observe real human behavior, not annotated datasets
- We experiment with tools that model stories, not labels
- We explore how emotions appear in leadership, power, conflict
- We test scenarios with actual context
- We study how AI amplifies or distorts emotional interpretation
- We allow for human contradiction and nuance
This requires slowing down. Observing. Testing. Iterating with people who care about behavioral truth, not just commercial applications.
The Long Game
I’m not chasing growth metrics or investor attention. EQ Engine is a slow, deep project – an experiment in how humans and AI might explore emotion together, honestly.
It starts small. A branch on my author site. A place where early experiments can happen quietly, where we can ask better questions:
How do leaders project emotion strategically? What blind spots do AI models have about human behavior? How can narrative and emotional modeling come together?
Eventually, EQEngine.io will stand on its own. But for now, it begins with people who sense the gap I see – and want to explore it deeply.
The world doesn’t need more emotional-AI hype. It doesn’t need more tools that label without understanding.
It needs a place where we can slow down and explore the truth of emotion – with all its contradictions, complexity, and human depth.
That’s what I’m building.
Not because I have all the answers.
But because I’m tired of asking the wrong questions.
