Shepherd Dynamics: Computational Frameworks for Conflict Prediction

This research compendium presents two interconnected computational frameworks for conflict prediction under the umbrella of "Shepherd Dynamics." Paper I (Nucleation Detection) introduces a variance inflection detector that identifies phase transitions by locating peaks in the second derivative of rolling variance, achieving F1=0.77 across six transition types with 100% accuracy on commitment-type "calm before the storm" dynamics. Paper II (Compression Dynamics) formalizes conflict potential as the symmetric KL-divergence between actors' compression schemes—their internal predictive models of reality—demonstrating significant correlation with conflict intensity (r=0.33, p<0.05) and strong temporal precedence (divergence leads conflict, r=0.67). Together, these frameworks provide both mechanistic explanation (WHY: divergence) and temporal detection (WHEN: inflection) for sociopolitical phase transitions. Core insight: Conflict is compression divergence. Transitions are variance inflection. Peace is alignment.

What does this mean in practice? Shepherd Dynamics turns the abstract question "will this situation explode?" into a measurable signal—weeks before headlines break. When two groups start compressing reality differently (seeing the same facts but reaching opposite conclusions), conflict becomes thermodynamically inevitable; our framework detects this divergence while de-escalation remains possible. When collective behavior enters the eerie calm that precedes phase transitions (markets before crashes, societies before revolutions, relationships before ruptures), our detector fires. Deployed at scale, this enables: (1) early warning systems for humanitarian organizations tracking civil unrest, (2) institutional risk dashboards for investors monitoring regime stability, (3) platform integrity tools for social networks detecting coordinated radicalization, and (4) diplomatic intervention timing for conflict mediators. The core unlock: instead of reacting to crises, you see them forming—like watching pressure build before an earthquake. Peace isn't the absence of disagreement; it's when groups still share enough compressed reality to negotiate. We can now measure that alignment, and know when it's failing.