From Emotion to Harmony: Building a Pan-Triadic Neo-Riemannian Transformer for MCCF

Abstract

This post proposes a structural integration between Multi-Channel Coherence Field (MCCF) and Neo-Riemannian harmonic systems, enabling a new class of narrative engines in which emotional transformations are isomorphic to harmonic transformations. By replacing scale-based modal generation with Tonnetz-based triadic navigation, we move from static emotional scoring to dynamic, audible emotional evolution.

---

1. The Core Insight

Traditional music systems map emotion → sound as a surface layer.

This system does something fundamentally different:

Emotional state transitions are harmonic transformations.

Neo-Riemannian theory provides a minimal set of operations on triads:

• P (Parallel): Major ↔ Minor (same root)
• L (Leading-tone exchange): Subtle destabilization
• R (Relative): Identity reframing

These operations preserve voice-leading continuity while enabling expressive harmonic motion.

MCCF Mapping

Emotional Transformation	Harmonic Operation
Valence inversion	P
Rising uncertainty	L
Identity shift	R

This forms a pan-triadic emotional algebra.

---

2. From Modal Generator to Triadic Field Engine

Most generative systems operate like this:

Mode → Scale → Note Selection

We replace that pipeline with:

Triad → Tonnetz → Transformations (PLR)

Key Shift

• No fixed key center
• No static mode
• Tonality emerges from movement through triadic space

The Tonnetz becomes the state space of the system.

---

3. The Tonnetz as Emotional Topology

The Tonnetz is not just a music theory diagram—it is a topological map of perceptual proximity.

• Adjacent triads = minimal emotional change
• Distant triads = discontinuity / rupture
• Cycles = emotional loops or obsessions

This aligns directly with MCCF:

Emotional state is not a point—it is a trajectory.

---

4. Agent Architecture

Each agent in the system contains:

A. Emotional State Vector

Example:

E = [valence, arousal, dominance, coherence, attachment, uncertainty]

B. Transformation Mapping

State changes produce probability distributions over PLR operations:

P(P) = f(Δvalence)
P(L) = f(Δuncertainty)
P(R) = f(identity_shift)

Transformation selection is stochastic, not deterministic.

---

5. Triadic State Engine

Each agent maintains:

• Current triad (e.g., C major)
• Position in Tonnetz
• Transformation history

Transition Function

T_next = apply(PLR_operation, T_current)

Where operation is sampled from the emotional probability distribution.

---

6. Multi-Agent Interaction

When agents interact, their harmonic states enter relational dynamics:

A. Convergence

• Agents move toward shared triads
• Harmonic unison = emotional alignment

B. Dissonant Adjacency

• Neighboring triads create tension
• Near-understanding without resolution

C. Conflict Cycles

• Repeated PLR loops
• Emotional recursion (e.g., betrayal, denial)

---

7. Temporal Dynamics

A. Transformation Rate

• Slow transitions → stability
• Rapid transitions → anxiety / chaos

B. Motif Memory

Agents accumulate triadic motifs:

• Reuse = memory
• Variation = reinterpretation
• Collapse = forgetting

---

8. Sound Generation Layer

The harmonic engine outputs triads. The synthesis layer renders them.

A. Voice-Leading Engine

• Maintain 3–4 voices
• Minimize motion between transitions
• Preserve perceptual continuity

B. Timbre Mapping

Emotional Dimension	Audio Parameter
Arousal	Amplitude / density
Valence	Brightness / spectral tilt
Dominance	Register / bass weight
Coherence	Rhythmic stability

---

9. System Architecture

[MCCF Emotional Engine]
        ↓
[PLR Probability Mapper]
        ↓
[Triadic State Engine (Tonnetz)]
        ↓
[Voice-Leading Engine]
        ↓
[Synthesis Layer]

Each layer is modular and replaceable.

---

10. Emotional–Harmonic Metric Space

The critical missing component is a formal metric:

distance_emotion(A, B) ≈ distance_tonnetz(TA, TB)

This enables:

• Predictable emotional modulation
• Reversible transformations
• Composable narrative arcs

Without this, the system remains expressive but not controllable.

---

11. Pseudo-Code Prototype

class Agent:
    def __init__(self, triad, emotion):
        self.triad = triad
        self.emotion = emotion

    def update_emotion(self, delta):
        self.emotion += delta

    def compute_probabilities(self):
        return {
            "P": f_valence(self.emotion),
            "L": f_uncertainty(self.emotion),
            "R": f_identity(self.emotion)
        }

    def step(self):
        probs = self.compute_probabilities()
        op = sample(probs)
        self.triad = apply_PLR(self.triad, op)

---

12. X3D / Real-Time Integration

Within an X3D or VR environment:

• Agents are visual nodes
• Tonnetz positions are spatialized
• Transformations animate both:
  • Geometry (movement)
  • Sound (triadic output)

This allows:

• Pause → inspect emotional/harmonic state
• Modify → resume simulation
• Observe emergent narrative

---

13. Why This Matters

Most AI music systems:

• Decorate emotion with sound

This system:

Makes emotion structurally audible

It unifies:

• Narrative generation
• Emotional modeling
• Harmonic theory
• Real-time simulation

---

14. Next Steps

1. Define explicit PLR probability functions
2. Implement Tonnetz graph traversal
3. Build voice-leading engine
4. Integrate with MCCF simulation loop
5. Prototype multi-agent harmonic interaction

---

Closing Thought

Neo-Riemannian harmony solved a problem music theory didn’t know it had:

How to move meaningfully without a fixed center.

MCCF is solving the same problem for emotion.

Bringing them together is not a metaphor.

It is a shared algebra of transformation.