Akihito Kimura A Data-Driven Approach to Music Composition, Sound Design, and Emotional Architecture
This article explores music composition as a system for designing emotional response, rather than merely expressing feelings. Drawing from over 30 years of experience, music theory, sound design, and data-driven audio analysis, it presents a framework for understanding how composers and producers can shape perception through structured constraints, probability, and state transitions. Whether you are working in electronic music, film scoring, or experimental sound, this perspective reframes composition as the architecture of human experience.
— A conclusion drawn from 32 years of practice and data science
Musicians are often described as “expressing emotion.”
But after more than three decades of composing, studying music theory across cultures, and analyzing sound through data, I have arrived at a slightly different conclusion.
Music is not the expression of emotion.
It is the design of conditions under which emotion inevitably emerges.
A composer does not directly manipulate the listener’s feelings.
Instead, we construct a system in which emotional transitions occur naturally.
I define my role as a State Transition Designer—
or perhaps more intuitively, a navigator of human consciousness.
Dominant Motion as the Placement of Instability
One of the first concepts taught in music theory is dominant motion:
the idea of resolving tension into stability.
In practice, however, this is not merely a rule—it is a matter of design.
The real question is:
When, and to what degree, should instability be introduced into the system?
When I was studying composition, my teacher once told me to draw shapes on paper before writing notes.
Not to arrange “sad sounds,”
but to design the precise moment when equilibrium begins to collapse—
the instant just before a tear falls.
This is not unlike writing a story.
You do not explain a character’s emotions directly.
You construct the situation in which those emotions must arise.
Music as a Constrained Probability System
Music exists between two domains:
- The objective world of physical phenomena (frequency, amplitude, spectrum)
- The subjective world of human perception and preference
Between these, music operates as a constrained probability system.
It is neither complete freedom nor rigid determinism,
but something in between.
Last year, I composed two full albums without touching a keyboard—
working entirely within a sequencer and piano roll.
Within the constraints of functional harmony, I controlled the entropy of sound,
guiding the system toward moments where a phase transition in perception occurs.
In this process, I rely on concepts such as:
- Hysteresis:
The meaning of a sound is determined by what came immediately before it. - Affordance:
The physical properties of sound—timbre, rhythm, texture—suggest specific perceptual responses.
These are not merely intuitive ideas.
They are supported, at least partially, by patterns extracted through data analysis.
Music Theory as an Optimization of Constraints
To me, music theory is not a set of rules.
It is a collection of constraint systems that have survived optimization across history.
This is precisely why I intentionally disrupt them—
by introducing non-Western scales, microtonal deviations, or alternative tunings.
Not to reject the system,
but to expand its probability space.
The moment something deviates from expectation—what we often call an “out” note—
the system’s search space widens dramatically.
And in that expansion,
the listener’s perception can access entirely new territories.
Why Ideas Do Not Run Dry
People sometimes ask me:
“How do you keep coming up with new music?”
For me, composing is not a special act.
It is closer to metabolism.
Like drinking water when you are thirsty.
What matters is not talent, but process.
In my case, that process can be described as follows:
- A large search space
Theory, physics, mathematics, psychology—each adds dimensions to how sound can be understood. - Continuous exploration
Small perturbations are constantly tested:
“What happens if I change this density by 1%?” - Delayed evaluation
Judgment is not applied during generation, but afterward.
In short:
I walk continuously through a vast search space,
and only later do I draw the map.
As long as this process is maintained,
ideas do not run dry.
Because they are not something I create.
They are something that flows.