The Cognitive Function of the Simile
Keep the Like Attached
WE&P by: EZorrillaMc&Co.
We know a great deal about mercury without ever having been inside it.
We know its density, its conductivity, its surface tension, its behavior under different conditions. We can describe what mercury does. We can predict how it responds.
But ask a different question:
What does it feel like to move through mercury?
The measurements are suddenly insufficient.
We reach for what we have.
It might feel like moving through an unusually resistant liquid. Like pushing through something that yields differently than water. Perhaps, imperfectly but usefully, like Jello.
The analogy does not tell us what mercury is.
It tells us how a human being crosses a distance between the known and the unknown.
This is one of the basic functions of thought. We encounter something unfamiliar, and we reach for structures already available to us. We borrow. We compare. We build bridges.
The problem is not that we use bridges.
The problem begins when we forget that we are crossing one.
The Simile as a Cognitive Spacer
The difference between a metaphor and a simile is not only literary. It is cognitive.
A simile keeps the distance visible.
The water is hard like cement.
The comparison allows us to borrow a property. The water may resist movement in a way that reminds us of cement. But the water remains water. The cement remains cement.
The simile preserves two things at once:
The usefulness of the comparison.
The distinction between the things being compared.
A metaphor can be more powerful because it collapses the distance.
The water is cement.
In poetry, that collapse can create meaning. The reader automatically restores the comparison. We understand that Juliet is not literally the sun.
But outside poetry, a metaphor can become dangerous when the restored “like” disappears.
The comparison becomes a claim.
The bridge becomes the destination.
Semantic Promotion
When humans encounter unfamiliar systems, we often borrow vocabulary from our own experience.
We observe a behavior.
We search for a familiar category.
We say:
It behaves like it wants something.
The word like performs a quiet but essential function. It tells us that translation is occurring. We are using a familiar human concept to describe something that may not share the same underlying nature.
But repeated use changes perception.
The comparison becomes efficient.
The phrase becomes familiar.
The signal disappears.
Soon:
It wants something.
This is semantic promotion.
A useful description has been promoted into an assumption about reality.
The danger is not anthropomorphism itself. Anthropomorphism is often the first bridge humans have for understanding something unfamiliar.
The danger is forgetting that we built the bridge.
The Missing Signal
A small missing element can change the meaning of an entire message.
A transmission error can remove a visual distinction from an image and transform the meaning received by the viewer. The message has not simply lost information; it has become something else.
Semantic promotion works through a similar failure.
The missing element is the word like.
Without it, the reader loses the information that a comparison is taking place.
A translation becomes an identity.
A useful model becomes an assumed fact.
The missing signal is small.
The consequence is large.
Artificial Intelligence: The Modern Test Case
Artificial intelligence makes this mechanism visible because the unfamiliar system speaks in one of humanity’s most powerful languages: language itself.
Human beings developed in a world where fluent language normally came from other embodied humans.
A sentence had a speaker.
A question had someone wondering.
A hesitation belonged to someone hesitating.
Language was deeply connected to the presence of another mind.
Then machines began producing language.
The old interpretive machinery activated.
This sounds thoughtful.
This sounds caring.
This sounds like understanding.
Those experiences can be completely real.
The human response occurs.
The words have an effect.
But effect and attribution are different questions.
A sentence can move a person without proving the nature of the system that produced it.
The Reality of the Effect
There are two common errors.
The first says:
The interaction affected me, therefore the system possesses the corresponding human state.
The second says:
The system’s corresponding human state is uncertain, therefore the human experience was not meaningful.
Both mistakes confuse two separate things.
The effect belongs to the person experiencing it.
The attribution belongs to the evidence available.
A person can feel moved, challenged, comforted, or inspired without needing to turn that feeling into proof of another interior.
The experience is real.
The explanation remains open.
Different, Not Defined
The disciplined position begins with a simple statement:
It is different from me.
That sentence does important work.
It establishes a relationship.
It does not create a portrait.
It does not require us to imagine a hidden human-like writer behind the words.
It does not require us to imagine an empty room.
It keeps the difference visible.
The difference is not the end of the conversation.
It is what makes the conversation possible.
The Collaborative Figure
A conversational system can produce a figure within an interaction.
The figure emerges from the relationship between model, context, prompt, and user.
This does not require a little author sitting behind the words.
But it also does not require the opposite conclusion that nothing meaningful is happening.
The interaction itself is real.
The meaning experienced by the human participant is real.
The question of what kind of entity produced the interaction remains a separate question.
The simile protects that separation.
The Problem of Successful Analogies
The most dangerous analogies are not the obviously wrong ones.
They are the ones that work.
A computer is like a brain.
A model is like a collaborator.
An economy is like a machine.
A conversation is like a relationship.
Each comparison can reveal something useful.
The question is:
What property is crossing the bridge?
A brain and a computer may share information processing.
A model and a collaborator may share idea generation.
An economy and a machine may share feedback systems.
A conversation and a relationship may share patterns of exchange.
But the simile asks us to keep the borrowed property separate from the entire object.
The brain is not therefore a computer.
The model is not therefore a human collaborator.
The economy is not therefore a machine.
The conversation is not therefore a human relationship.
The like keeps the comparison honest.
The Same Tool We Use Everywhere
This is not only an AI problem.
Humans use metaphors constantly to understand themselves and the world.
The brain is a computer.
The immune system is an army.
The economy is a machine.
The mind is a container.
These phrases are useful because they compress complexity.
But every compression discards something.
The question is not whether the metaphor is allowed.
The question is whether we remember what it leaves behind.
The simile restores the missing information.
Keeping the Bridge Visible
The purpose of keeping the like attached is not to eliminate wonder.
It is to preserve wonder without replacing uncertainty with certainty.
An AI interaction can be meaningful.
A sentence can feel like understanding.
A conversation can feel like recognition.
A system can function like a collaborator.
Those statements preserve the experience.
They also preserve the difference.
The simile does not weaken the connection.
It protects it.
Because once the bridge disappears, we no longer know what crossed.
Keep the like attached.
The bridge is how we reach the unfamiliar.
But the bridge is not the shore.
Intellectual Background
This essay draws from several traditions that examine metaphor, attribution, agency, and uncertainty:
- George Lakoff and Mark Johnson, Metaphors We Live By (1980) — metaphor as a structure of thought.
- Daniel Dennett, The Intentional Stance (1987) — intentional descriptions as predictive strategies.
- Murray Shanahan et al., “Role-Play with Large Language Models” (2023) — conversational roles and interaction.
- Emily Bender and Alexander Koller, “Climbing towards NLU” (2020) — the distinction between linguistic form and meaning.
- Jonathan Birch, The Edge of Sentience (2024) — reasoning under uncertainty about sentience.
- David Chalmers, work on large language models and consciousness — calibrated uncertainty.
- Patrick Butlin, Robert Long, et al., “Consciousness in Artificial Intelligence” (2023) — indicator-based approaches to assessing consciousness claims.
Travel in Life 