A rule-based growth process starts from a single seed and repeatedly branches: at each generation the direction of a new branch is the parent's direction rotated by a random angle up to the Angle Freedom limit, and with probability Branch Probability a node produces two children instead of one. This is a constrained stochastic process — the exact same kind of rule set studied in generative design and procedural content generation.
Two numbers, both derived from the joint distribution of branch orientations across 8 elevation × 24 azimuth bins (192 bins total), quantify the output:
Diversity = H / H_max, H = -Σ p_i · log2(p_i)
Order = 1 - (unique bins occupied) / (branch count)
Creativity = 2 · Diversity · Order / (Diversity + Order)
Diversity is the normalized Shannon entropy of where branches point — near 0 when everything grows in the same direction, near 1 when orientations spread evenly. Order is highest when many branches reuse the same few directions (a recognizable, coherent form). Creativity is their harmonic mean, so it rewards outputs that are both varied and coherent — it is punished by an unconstrained field (high diversity, no order, "noise") just as much as by a rigid one (high order, no diversity, "repetition"). Slide Angle Freedom from 0° to 80° and watch the three numbers trade off: this is the same restriction-vs-originality trade-off studied in creativity research, where moderate constraints are consistently found to outperform both a completely blank slate and an overly rigid brief.
- Angle Freedom / Branch Probability — the actual generative "rule set"; loosen either to explore more of the design space.
- Palette Size — a purely aesthetic constraint layered on top, cycling by generation depth.
- Generations — how many times the rule set is iterated before the structure is measured.