When I first wrote a slightly chaotic document exploring these ideas, I sent it to someone close to me who helped me think deeply about these questions and extend them. What began as a private reflection evolved into a broader framework for thinking about emotion and logic in AI — a conversation that ultimately gave me the courage to share it publicly.
As AI proceeds to shape our choices and relationships, understanding what makes human intelligence whole has never been more important.
This is a three-part series exploring how neuroscience reveals why current AI is fundamentally incomplete and how we could build something better.
About This Series
Part 1: When the Mind Feels Nothing - What It Means for the Machines We Build (You’re reading it!)
Elliot’s story and what it teaches us about how the brain actually works. Why emotion isn’t the enemy of logic, but it’s infrastructure.
Part 2: Current AI’s Limitations
How today’s LLMs are essentially “logical Elliots” - brilliant at thinking but paralyzed when choosing requires feeling.
Part 3: Framework and Practical Applications
A technical deep-dive into how we could build emotionally-aware AI systems, with concrete examples and ethical considerations.
Introduction
We usually think intelligence is about knowing - about solving crazy math problems and chasing perfect answers. We rarely add caring to the equation.
So one day I was sitting and thinking that maybe the key to building better AI doesn’t lie in making it smarter, but in understanding why intelligence without emotion collapses. And I remembered the story - the proof that comes from a man who lost the ability to feel - and with it, the ability to live.
Meet Elliot: brilliant, logical, and completely dysfunctional. It’s an unsettling idea, that a person can lose feeling yet keep brilliance intact. That logic alone can’t hold a life together.
After having a tumor removed from his brain, Elliot’s IQ remained intact. His memory functioned perfectly. His reasoning was sharp. Yet his life fell apart. He couldn’t make decisions. He’d spend hours deciding between cereal brands, paralyzed by trivial choices. He lost his job. His relationships dissolved. Not because he lacked logic, but because he couldn’t feel.
This case, studied by neuroscientist Antonio Damasio, shattered the myth that emotion and reason are separate (see Damasio’s Descartes’ Error, which first described Elliot’s case). It revealed something profound: emotion isn’t the enemy of logic - it’s its essential partner. Without emotion, logic is incomplete. Intelligence cannot act.
And here’s the thing we’ve all experienced: when you’re stuck between choices - job offers, relationships, even which restaurant to pick - you’re essentially Elliot for a moment. Logic lists the pros and cons, but what we actually need is to stop and feel to decide. Aren’t we? ;)
The Neuroscience: Emotion as Infrastructure
What Emotion Actually Is (And It’s Probably Not What You Think)
Emotion is often misunderstood. We think of it as this vague, subjective feeling - something that happens to us but isn’t really us. We’re taught that emotions cloud judgment, that they’re unreliable, that logic is superior.
But that’s not what the neuroscience shows.
Let me propose a definition: Emotion is the process of routing signals that connect what we think to how our body reacts, using electrical impulses with different intensities directed to specific regions.
This might sound technical, but it reveals something crucial: emotion isn’t separate from cognition. It’s part of it. It’s signal, not noise. It’s infrastructure.
How does this differ from intellect?
Intellect is a combined process involving several systems working together. It’s the ability to read all incoming data: what you see, what you feel physically, emotionally, and to blend that with memory, knowledge, experience, and the search for new information. All of that gets used to create something new or solve problems.
So intellect isn’t separate from emotion. Intellect is an abstraction - and emotion is one of its core components.
This brings me back to Jeff Hawkins and his book “On Intelligence” (Hawkins, 2004). His formula is: Intellect = Memory + Prediction. Wait—he didn’t mention emotions at all. So is his formula flawed?
No. It’s not flawed. It’s incomplete.
Hawkins wasn’t trying to build a full cognitive architecture. He focused on sensorimotor prediction, particularly modeling cortical columns—how we predict patterns based on memory and incoming data. That doesn’t make his work wrong. But it does mean it doesn’t cover the whole story. The emotional component is missing—the component that turns prediction into action.
Why Logic Alone Fails: The Neuroanatomy
Let’s explore where logic “Lives”, shall we?
Logic isn’t one process in the brain. It involves multiple regions:
- Dorsolateral Prefrontal Cortex (dlPFC): This is where we hold multiple variables in mind, simulate if-then scenarios, plan ahead. It’s the “chess master” of the brain - slow, controlled, symbolic.
- Parietal Lobes: Especially active in formal logic, geometry, and coding, where visual-spatial structure matters.
- Left Temporal Lobe: This is where verbal logic gets parsed, where we build arguments and comprehend structured information.
Elliot’s logic centers were intact. He could think. He could calculate. He could reason through complex problems.
But he couldn’t choose. Let’s explore what actually happened to Elliot. Where, exactly, in the brain did things break down?
Where Emotion Integrates
The breakthrough came when neuroscientists realized that decision-making requires integration across systems:
- Amygdala: Emotional intensity, danger/safety signaling
- Insula: Interoception - feeling what’s happening inside the body
- Anterior Cingulate Cortex: Reading bodily signals and integrating them
- Ventromedial Prefrontal Cortex (vmPFC): This is the crucial region. Modern neuroscience shows the vmPFC isn’t a single uniform zone - it’s functionally heterogeneous, with different subregions handling emotion regulation, value representation, and social cognition (Wallis, 2007; Roy et al., 2012). But broadly, this region integrates cognitive reasoning with emotional signals and somatic (body-based) feedback. When damaged, it leads to exactly Elliot’s symptoms - intact logic but emotional-decision paralysis (Damasio, 1994).
This is where Elliot’s damage occurred.
The Integration Point: The Interoceptive Loop
Here’s what happens in a healthy brain when you make a decision:
- The autonomic nervous system activates (heart rate changes, gut tension, sweat)
- Interoception kicks in: The insula reads these bodily signals - this is called interoception, the brain’s map of your internal state (Craig, 2002). The anterior cingulate cortex then integrates these signals with emotional context.
- The vmPFC weighs all of this - bodily sensations, emotional significance, and logical considerations - to form value judgments (Hare et al., 2009; Tusche et al., 2010)
- You get a final, weighted decision: not just “what is,” but “what matters”
Modern research emphasizes that this isn’t a linear pipeline but a dynamic network where emotion, cognition, and body signals continuously interact (Sokol-Hessner & Rutledge, 2019). And you can see why this matters: your brain literally needs to feel your gut tension or your racing heart to make choices that align with what your body knows.
So even our “purest logical thought” isn’t purely logical. It’s constantly being colored, prioritized, and modulated by bodily states.
This means that logic can:
- Tell you what is consistent, but not what is good or urgent or meaningful
- Identify patterns, but not decide if they matter to you
- Propose many solutions, but can’t tell you which one feels right to act on
Evidence from lesion studies suggests the vmPFC’s causal role in integrating these signals (though researchers are still mapping exactly which subregions control which functions) (Noonan et al., 2012; Hunt et al., 2012). But the pattern is clear: when this integration breaks down, decision paralysis sets in.
Elliot proved this devastatingly. He could analyze every option logically, but without emotion, he couldn’t select. Everything felt equally important - or equally trivial.
Decision-Making as Collaboration
The real insight is that decision-making isn’t a battle between logic and emotion. It’s not even a partnership. It’s a negotiation - between logic, emotion, and the body.
Let me walk you through a real example from my own experience while writing the first draft of this:
I wake up at 6:39 AM after not sleeping all night (yeah, try to understand this logically!). I want coffee. But I feel fear in my stomach. The logical parts of my brain say: “You should sleep. You haven’t slept. Coffee will make this worse.” But another part wants coffee. And then a third part suggests: “Write instead. This exploration feels exciting.”
Here’s what was actually happening:
- Coffee desire → Impulse/desire (initial action bias)
- Feel in the stomach → Somatic check (fear signal)
- It’s 6:39 AM → Rationality kicks in (time awareness)
- I wasn’t sleeping → Rational fact (state awareness)
- I want coffee anyway → Emotional feedback loop
- I’ll write instead, it feels exciting → Reframing with synthesis
- I made a decision → Commitment and emotional closure
This isn’t a linear process. It’s not logic versus emotion. It’s like three people talking to each other until they align. Logic. Emotion. Body. They’re not in conflict. They’re in negotiation. They’re trying to argue until they reach alignment - not agreement, not “truth.” Just alignment. A shared “yes.”
This is how healthy decision-making works. But it requires all three systems.
Somatic Markers: Emotional Shortcuts
Here’s where things get really interesting. Every time you make a decision, your brain remembers how it felt last time. That memory is stored not just cognitively, but emotionally - as a pattern of bodily activation (tension, nausea, excitement, warmth).
Then, the next time you face something similar, your brain reactivates that feeling even before you consciously think about it.
These are called somatic markers - a concept Damasio developed from studying cases like Elliot (Damasio, 1994). They act like emotional shortcuts.
A girl eats ice cream while experiencing a traumatic event. Even years later, she might avoid ice cream. Not because she remembers the trauma cognitively - but because her body remembers. The feeling gets tied to the experience. So now, ice cream = dread. This is a well-known phenomenon in therapy. It’s not logic. It’s not memory. It’s somatic resonance.
Music works similarly. Sometimes when I hear a song, I don’t remember the scene first. I feel it first instantly, and then I recall the situation (sometimes I need to recall it for a very long time). The whole emotional landscape floods back. It’s not a snapshot. It’s an echo in the body.
And this system can backfire. Sometimes, we feel safe in situations that are actually unsafe. Why? Because if a person never learned what safety feels like, the unfamiliar - even if it’s safe - triggers fear. And the familiar - even if it’s dangerous - feels “right.” Not because it is, but because it matches what our body knows.
Emotion isn’t just an influence. It’s infrastructure. It helps us decide. It helps us filter. But sometimes, it also traps us if our inner compass was broken on life’s journey.
The Decision-Making Loop
When all three systems - logic, emotion, body - align, decisions flow. But when they don’t?
That’s when people get stuck and overwhelmed. That’s when you lie awake at night, weighing options that all seem equally valid or equally impossible. We’ve all been there - and now we know why it’s so hard.
Sometimes, a decision needs completion, but none of the options feel right. The systems can’t align. No resonance is found. So the brain loops - trying, again and again, to reduce discomfort in the somatic system.
If there’s no resolution, no good options, just the weight of having to choose - that’s why it hurts. That’s why it loops.
And that’s why we can’t know what we’d do in situations we’ve never been in. Because it’s not just logic that decides. It’s emotion + memory + body + context. We might think we’d never lie, or that we’d always follow our ethics, or die for someone we love. But the truth is - we won’t know until the moment happens.
Until the full system engages - until all three voices speak - we’re just guessing.
This phenomenon where people struggle to accurately imagine how they’d feel or act in a situation they’ve never experienced - even when they think they’re being reasonable - is called the “Hot-Cold Empathy Gap” (Loewenstein, 2005). It’s a well-studied cognitive bias.
Which brings us to empathy. Maybe empathy isn’t about understanding how another person feels. Maybe it’s about understanding that sometimes we just can’t understand. Because we haven’t been there. Because our system hasn’t processed that configuration of fear or grief or shame or love.
So real empathy isn’t saying, “I get it.” It’s saying, “I know I might never fully get it. But I see that you do. And I believe you.”
And that’s enough.
A Crucial Distinction Before We Bridge to AI
When we explore computational analogues for emotion-like processing in Parts 2 and 3, we must be clear—these are behavioral simulations, not biological replications. Machines won’t “feel” emotions in the way humans do. They won’t have somatic markers, interoceptive loops, or the subjective experience of fear, joy, or dread.
What we’re exploring is whether computational systems can simulate the patterns of emotional-weighting and value-signaling that make human decision-making possible. Not feeling. Not consciousness. Pattern recognition and routing.
Elliot’s paralysis came from the absence of biological emotion. AI’s paralysis comes from the absence of computational analogues. Different mechanisms, similar outcome.
A brief technical note: What I’ve called “emotion” throughout this piece encompasses overlapping processes—emotion (reactive value-signaling), motivation (goal-directed allocation), and affect (valenced states). The neuroscience I’ve drawn on treats these as integrated rather than modular: not three separate agents “negotiating,” but one hierarchical system (predictive coding, active inference) where valuation happens at every level. When Parts 2 and 3 explore computational analogues—somatic marker equivalents via reinforcement learning, Bayesian priors, or affective computing models—we’ll distinguish behavioral simulation (pattern weighting) from intrinsic motivation (genuine valuing). The question isn’t whether machines can feel, but whether they can simulate the computational patterns that make prioritization and choice possible.
Conclusion: The Foundation for What’s Next
Elliot’s story leaves us with a simple, unsettling truth: logic alone can think, but it cannot choose. It can model the world, but it cannot care about it.
Emotion isn’t a luxury - it’s the pulse that gives reason direction. The mind doesn’t work through separation but through connection: thought, feeling, and body speaking in the same language of meaning.
When that dialogue breaks, as it did for Elliot, intelligence becomes weightless. It can analyze everything, yet move toward nothing.
In Part 2, we’ll turn this mirror toward machines. We’ll see how today’s AIs, for all their brilliance, resemble Elliot - perfect logic suspended in emotional vacuum, able to calculate but unable to value.
And in Part 3, we’ll explore how to change that - how emotion-like architectures could help machines not replace human judgment, but support it, learning to weigh context, care, and consequence.
We can’t build great intelligence that understands us until we remember what it means to be us.
References
- Craig, A. D. (2002). “How do you feel? Interoception: the sense of the physiological condition of the body.” Nature Reviews Neuroscience, 3(8), 655–666.
- Damasio, A. R. (1994). Descartes’ Error: Emotion, Reason, and the Human Brain. New York: G.P. Putnam’s Sons.
- Damasio, A. R. (1999). The Feeling of What Happens: Body and Emotion in the Making of Consciousness. New York: Harcourt.
- Hare, T. A., et al. (2009). “Self-control in decision-making involves modulation of the vmPFC valuation system.” Science, 324(5927), 646–648.
- Hawkins, J. (2004). On Intelligence: How a New Understanding of the Brain Will Lead to the Creation of Truly Intelligent Machines. New York: Times Books.
- Hunt, L. T., et al. (2012). “Mechanisms underlying cortical activity during value-guided choice.” Nature Neuroscience, 15(3), 470–476.
- Loewenstein, G. (2005). “Hot-Cold Empathy Gaps and Medical Decision Making.” Health Psychology, 24(4S), S49-S56.
- Noonan, M. P., et al. (2012). “Separate value comparison and learning mechanisms in macaque medial and lateral orbitofrontal cortex.” PNAS, 107(47), 20547–20552.
- Roy, M., et al. (2012). “Ventromedial prefrontal-subcortical systems and the generation of affective meaning.” Trends in Cognitive Sciences, 16(3), 147–156.
- Sokol-Hessner, P., & Rutledge, R. B. (2019). “The computational and neural substrates of moral strategies in social decision-making.” Nature Communications, 10, 1483.
- Tusche, A., et al. (2010). “Automatic processing of political preferences in the human brain.” NeuroImage, 49(1), 914–923.
- Wallis, J. D. (2007). “Orbitofrontal cortex and its contribution to decision-making.” Annual Review of Neuroscience, 30, 31–56.