How Cells Decide

You cut your hand slicing bread this morning. Nothing serious. You rinsed it, pressed a paper towel to the wound, and forgot about it. But under the skin, a conversation started.

Platelets arrived first, clotting the breach. They released signaling molecules that summoned white blood cells. The white blood cells released their own signals, calling fibroblasts to begin rebuilding the extracellular matrix. Epithelial cells at the wound’s edge began migrating, then dividing, crawling across the provisional scaffold, knitting the gap closed.¹ As inflammation resolved and the tissue re-epithelialized, pro-growth cues declined and contact-dependent brakes kicked in.²

No one coordinated this. No central office dispatched the platelets or told the fibroblasts when to stop weaving. Each cell responded to its local environment: the molecules it sensed, the neighbors it touched, the mechanical tension in the tissue around it. The repair emerged from a thousand local decisions, none of which contained the plan.³

A week from now, the cut will likely be gone.⁴ You will not have thought about it once. But your cells thought about it constantly. They computed, communicated, decided, and stopped when the job was done.

If this counts as intelligence at all, it is intelligence made of relationships, not of a central controller. The question is whether that framing holds up, and what it means if it does.

The cognitive loop

Consider a bacterium swimming through murky water.⁵

It needs to find food and avoid toxins. This seems like a simple problem until you list what it requires. The bacterium must sense the chemical concentration around it. It must remember what the concentration was a moment ago. It must compare past and present. It must decide whether conditions are improving or deteriorating. And it must act on that decision, either continuing in its current direction or tumbling randomly to try a new one.

Sense, remember, compare, decide, act. That sequence is a cognitive loop. It is happening in an organism with no neurons, no synapses, no brain. At roughly two microns long, the bacterium is only a few wavelengths of visible light across.⁶

The mechanism is chemistry. Receptors on the bacterium’s surface bind to molecules in the water. That binding triggers a cascade of protein modifications inside the cell. The cascade affects proteins that control the flagellar motor, the little propeller that moves the bacterium through liquid. Crucially, these modifications happen on a timescale that creates memory: the cell’s internal state reflects not just current conditions but recent history. In control theory terms, it behaves like an integral feedback system, resetting its baseline so the cell stays sensitive to changes across a wide dynamic range.⁷

What’s happening here is computation. Not metaphorically. The bacterium is processing information, storing state, and executing biased switching: fewer tumbles when things improve, more when they worsen. The implementation is protein interactions rather than silicon transistors. But the logic is recognizable: state, comparison, conditional action.⁸

The standard story says intelligence requires a brain. Neurons are the hardware of thought; more neurons, more intelligence. This story has a problem. The bacterium, by any functional measure, is making decisions. It has zero neurons. I am using cognitive language here as a functional description, not a claim about consciousness.⁹ But if the function fits, either we need a different definition of decision-making, or we need a different theory of what makes it possible.

Becoming

The deeper puzzle is not that cells compute. It is that they decide what to become.

You began as a single fertilized egg. That egg divided, and divided again, and eventually became the tens of trillions of cells that compose your body.¹⁰ Every one of those cells contains the same DNA, the same instruction manual. Yet they became radically different things: neurons that fire electrical signals, muscle fibers that contract, red blood cells that carry oxygen, the transparent crystallin-packed cells of your eye’s lens.¹¹

Same instructions. Different outcomes. How does a cell know which outcome to choose?

The answer is context. Cells do not decide in isolation. They decide based on where they are, who their neighbors are, and what signals those neighbors are sending. Development is a conversation, not a blueprint.¹²

Picture an early embryo. A cell somewhere in that cluster is receiving signals from its neighbors: molecules that bind to receptors on its surface and activate specific genes inside. Those newly active genes change what the cell produces, including what signals it sends back to its neighbors. The neighbors respond by adjusting their own gene expression. The whole system evolves together, each cell’s fate shaped by its position in the network of conversations.¹³

What emerges is distributed decision-making. No single cell contains the master plan. The plan is not located anywhere. It emerges from the interactions. The intelligence is relational.¹⁴

Landscapes and attractors

Inside each cell, genes form networks. Gene A produces a protein that activates Gene B and suppresses Gene C. Gene B’s product activates Gene D. Gene D’s product suppresses Gene A. Thousands of genes interlock in loops of activation and repression.¹⁵

These networks have stable states. Certain combinations of active and inactive genes reinforce themselves: the active genes keep each other on, the inactive genes keep each other off. A cell “decides” to become a heart cell when its gene network settles into the heart-cell configuration. Once there, it tends to stay, often for years or a lifetime, because feedback stabilizes the state. But strong perturbations can still push cells across ridges into other basins.¹⁶

The biologist Conrad Waddington visualized this as a landscape.¹⁷ Imagine a ball rolling down a hillside carved with valleys. The ball can roll into any valley, but once it settles at the bottom, it tends to stay. The valleys are the stable cell types: neuron, muscle, skin. The ridges between them are the barriers that prevent a cell from switching identities. The ball’s path down the landscape is development.

It is a metaphor with real mathematics behind it: gene-regulatory dynamics can be analyzed as flows through a high-dimensional state space, where cell types correspond to attractors. In many cases we can even construct an effective “landscape” from the dynamics, though it comes with important caveats about when such a global potential function actually exists.¹⁸

The developmental biologist is mapping these basins. Some are shallow: easy to enter, easy to leave. Some are deep: once a cell falls in, it almost never escapes. Cancer, in this view, is a cell finding a new attractor that evolution never intended, a valley that should not exist.¹⁹

The wisdom of the swarm

Individual cells impress. Collectives astonish.

Many bacteria practice quorum sensing.²⁰ They continuously release signaling molecules into their environment. When the concentration of these molecules crosses a threshold, indicating that enough bacteria are nearby, the entire population switches behavior simultaneously. They might activate virulence genes, form a biofilm, or begin producing light.²¹

In effect, this is collective decision-making. Each bacterium is asking: Are there enough of us? No single bacterium knows the total count. They each respond to the local concentration of the shared signal. The population-level answer emerges from individual responses to a common medium. The decision is made by no one and everyone.²²

The slime mold Physarum polycephalum goes further.²³ Despite being a single cell, it can grow to cover several square feet, a pulsing yellow network spreading across a forest floor or a laboratory petri dish. When searching for food, it extends tendrils in all directions. When it finds food sources, it reinforces the paths between them and withdraws from dead ends. The result is a network that efficiently connects all the food.

Researchers placed oat flakes on a map of the Tokyo metropolitan area, each flake representing a major city. The slime mold, starting from “Tokyo,” grew a network connecting the flakes. The resulting network was strikingly similar in layout, and comparable to the Tokyo-area rail system on measures like efficiency and fault tolerance.²⁴ It had solved, through local growth rules, an optimization problem that human engineers solved with computers and deliberation.

No central planner. No global map. Just: reinforce what works, retract what doesn’t. From this simple rule, near-optimality emerges.²⁵

What the hardware theory misses

The standard theory of intelligence is hardware-centric. You need neurons. You need a brain. The brain is the seat of cognition; the body is its vehicle. Complexity of thought scales with complexity of neural architecture.²⁶

This theory has explained much. It fits the broad pattern that animals with larger and more complex brains tend to exhibit more sophisticated behaviors. It helped link psychiatry to neuroscience. It gave us the metaphor of the brain as computer, which, whatever its limitations, has been enormously productive.²⁷

But cells complicate the story.

Here we have sensing, memory, comparison, decision, and action, all implemented in molecular networks with zero neurons. We have developmental programs that reliably construct complex organisms through distributed computation. We have collectives that solve optimization problems human engineers struggle with.

None of this denies that neurons are extraordinary. Fast, energy-efficient electrical signaling is a powerful specialization. But specialization is not a prerequisite for feedback-driven decision-making. Cells were making decisions for billions of years before neurons existed.²⁸

The key ingredients are not specific molecules or cell types. They are relationships. Sensing: components that respond to their environment. Memory: persistence of state over time. Communication: information flow between components. Feedback: responses that influence future behavior. Stability: patterns that maintain themselves once established.²⁹

Assemble these in the right configuration and you get something that looks like decision-making. Whether you call it intelligence is a semantic choice. Functionally, it is hard to distinguish from cognition.³⁰

Intelligence is organizational, not material. It is about how parts relate, not what parts are made of. Neurons are good at certain kinds of processing, but they are not the only substrate for adaptive response. They are one solution among many.

The democracy of the body

Your body is not a dictatorship with the brain as absolute ruler. It is something more like a democracy: a distributed system where countless local decisions aggregate into coherent action.

Your gut microbiome can influence brain function through immune, endocrine, and neural pathways, though the strength and direction of effects vary across contexts and individuals.³¹ Your immune system makes sophisticated judgments about what counts as self and what counts as foreign, a distinction more nuanced than most philosophy.³² Your heart contains an intrinsic network of neurons that helps regulate and coordinate cardiac function, though the basic beat is generated by pacemaker cells in the sinoatrial node, even without brain input.³³

The brain is important. But it is less a commander than a coordinator. One voice in a chorus of cellular conversations. A particularly loud voice, perhaps. But not the only one speaking.³⁴

This has implications that ripple outward. If cognition is distributed, where does the self reside? If cells make decisions, are they morally considerable? If intelligence is organizational, how should we think about the boundaries of mind?³⁵

I find these questions more interesting than the old debate about whether machines can think. Cells make it hard to deny that many cognitive functions don’t require brains. The question is what kinds of organization produce what kinds of thought. And what we owe to the systems that think in ways we are only beginning to understand.

The patience of biology

One more lesson from cells: the power of patience.

Biological systems do not optimize for speed. They optimize for robustness, adaptability, and long-term survival. Evolution takes millions of years. Development takes months. Wound healing takes weeks. By engineering standards, these timescales are glacial.³⁶

But the results are remarkable. Self-repairing systems. Adaptive immunity that remembers threats for decades. Bodies that maintain themselves for a century. These achievements exist not despite the slowness but because of it. The long timescales allow for error correction, redundancy, graceful degradation. The kind of deep reliability that fast systems cannot achieve.³⁷

We build systems that optimize for speed and then wonder why they are fragile. Cells optimize for persistence and achieve both reliability and adaptability. There is a lesson here about complex systems of any kind. Sometimes the fast answer is the wrong answer. Sometimes the distributed solution outperforms the centralized one. Sometimes intelligence is not in the individual but in the interaction.

Cells figured this out three billion years ago. By the time you remember the cut on your hand, the democracy beneath your skin will have already adjourned.