I recently completed a new research paper titled Beyond the Cognitive Horizon: Formalizing the Recursive Systems Causal Decision Model and Its Extension to Recursive Metaphysical Dynamics.
DOI: 10.5281/zenodo.21879539
The paper started with a question that sounds simple:
How far has humanity advanced because of what the human mind is capable of understanding, and what happens when the systems we create become too complex for one human mind to fully comprehend?
That question became much larger than I expected.
For most of human history, our intellectual boundaries were tied to our biological boundaries.
We could only observe so much.
Remember so much.
Calculate so much.
Compare so many variables.
Hold so many relationships in our minds at once.
We learned to extend those limits through writing, mathematics, science, libraries, universities, computers, and institutions. Human civilization became a kind of distributed intelligence. No single person understands every part of a modern aircraft, computer network, hospital, electrical grid, economy, or scientific discipline. Yet collectively we continue to build them.
Artificial intelligence adds something new.
AI does not simply store more information. It can examine relationships across spaces that would be impossible for one person to search manually.
Recent advances in mathematics helped push me toward this question. AI systems are already assisting mathematicians in finding patterns, exploring formal proofs, and searching enormous mathematical spaces.
That creates an interesting distinction.
A machine might calculate something that a human cannot calculate manually.
It might identify a relationship that no human noticed.
It might even produce a valid proof.
But does that mean the machine understands the result in the same way a mathematician does?
And an equally important question follows:
Does the human have to completely understand the machine's internal reasoning before the knowledge becomes useful?
Those questions led me to what I call the Cognitive Horizon.
Every observer has limits.
A person can understand a system while its complexity remains within that person's ability to integrate its important relationships.
Eventually, however, a system can contain too many variables, interactions, feedback loops, delays, probabilities, and competing explanations to hold together mentally.
At that point, the system has crossed the observer's Cognitive Horizon.
This does not mean the system has become irrational.
It means the observer has reached a limit.
Something that looks chaotic from one level of understanding may appear structured from another.
A novice looks at a page of advanced mathematics and sees symbols.
A mathematician sees relationships.
A mechanic hears an engine and notices a failing component where another person hears noise.
The world did not change.
The observer's ability to recognize structure changed.
Artificial intelligence could expand that horizon.
But it creates another problem.
AI has limits too.
So a human plus AI does not produce unlimited intelligence.
It creates another observer with another horizon.
The second idea in the paper is what I call the Firmament Boundary.
The Cognitive Horizon concerns what you are capable of understanding.
The Firmament Boundary concerns what you are capable of accessing.
There is always information you can observe.
Information you can infer.
Information that exists but remains hidden.
And information you do not even know to look for.
That distinction matters because more information does not automatically produce better understanding.
You can have enormous amounts of data and still misunderstand what causes what.
Which led to another deceptively simple example.
Imagine that every time I light my gas stove, I happen to be wearing a red shirt.
After hundreds of observations, the data might show an almost perfect relationship:
Red shirt appears.
Stove lights.
Red shirt appears.
Stove lights.
A computer looking only for correlations might find the relationship immediately.
But the red shirt does not light the stove.
Fuel, oxygen, and ignition produce the flame.
Take away the red shirt and the stove still lights.
Take away the ignition and my beautiful red shirt accomplishes nothing.
That ridiculous example points toward a serious problem in management, science, artificial intelligence, and everyday reasoning.
Correlation is not causation.
Finding a pattern is not the same thing as understanding the mechanism producing the pattern.
That distinction became central to the new model.
The Recursive Systems Causal Decision Model, or RSCDM, examines how a decision-maker moves through a system:
Observe.
Interpret.
Identify patterns.
Generate explanations.
Form a hypothesis.
Make a prediction.
Intervene.
Observe what happens.
Learn.
Revise the model.
Act again.
The last part matters.
The manager or observer is not standing outside the system.
The decision changes the system.
The changed system produces new information.
That information changes the observer.
The observer makes another decision.
The process is recursive.
This means decisions should not be understood as isolated events.
They are part of continuing feedback between the observer and the world being observed.
One conclusion became especially important while developing the paper:
AI assistance is not automatically better than human reasoning.
Sometimes AI expands the available search space and helps identify relationships that a person missed.
Sometimes it produces convincing but incorrect recommendations.
Sometimes humans reject correct machine advice.
Sometimes humans trust incorrect machine advice because it sounds authoritative.
So the important question is not:
"Is AI smarter than humans?"
A better question is:
What does the human contribute, what does the AI contribute, and how do we test the combined system before acting on its conclusions?
That gives us another cycle:
Human reasoning.
AI analysis.
Causal testing.
Verification.
Action.
Observation.
Revision.
The AI should expand the investigation, not become an unquestioned oracle.
The paper contains a synthetic experiment using 240 simulated decision-makers across 1,920 decision trials.
Those are simulated cases, not human research participants.
The purpose was to see whether the proposed research design could produce measurable differences when variables such as complexity, information access, causal structure, and AI-task fit were deliberately changed.
The model did.
That is useful, but it is not the end of the research.
The next stage is much more important:
put the theory at risk with actual human participants.
If the predicted relationships fail, the theory must change.
That principle matters to me.
A theory should not survive because its creator keeps finding ways to explain away contradictions.
The theory should survive because it continues to work when somebody tries to break it.
This research also opened a larger door.
RSCDM deals with systems we can test.
Management decisions.
Information.
Human reasoning.
AI assistance.
Causation.
Feedback.
But the architecture raises a broader philosophical possibility.
What if similar structures occur at larger levels?
Objects exist within systems.
Systems interact through fields and relationships.
Constraints determine which changes are possible.
Interactions produce transformations.
Transformations cross thresholds.
Threshold events combine.
Clusters form.
New wholes emerge.
And those wholes become components inside still larger systems.
That larger framework became Recursive Metaphysical Dynamics, or RMD.
I am careful in the paper not to present RMD as experimentally established physics.
It is a metatheoretical extension.
RSCDM is the part we can put on the laboratory table.
RMD asks where the same principles might lead if they continue upward.
And that leads back to the question that started everything.
A human encounters something beyond human understanding.
AI helps the human reach farther.
Then AI encounters something beyond its own ability to represent.
What happens next?
Perhaps intelligence does not have a final horizon.
Perhaps every horizon simply reveals another.
The central idea of the paper eventually became two rules:
Do not mistake the edge of observation for the edge of reality.
Do not mistake the edge of present understanding for the final structure of what can be known.
That is where this research begins.
I have a feeling it is going to take me much farther than I originally intended.
Francisco M. Martinez
August 2026
Beyond the Cognitive Horizon: Formalizing the Recursive Systems Causal Decision Model and Its Extension to Recursive Metaphysical Dynamics
DOI: 10.5281/zenodo.21879539
