Detects unexplored concepts in AI creativity generation, indicating new idea pathways.
A creative generator is usually a black box: it produces ideas but cannot say what it fails to produce. Building on the Concept Collider—a glass-box engine in which every idea is a collision of two distant concepts along a fracture (a structural tension such as an internal contradiction or a hidden dependency)—we map not only what a generator makes but its negative space: what it structurally fails to make. Because every output is traced to its fracture, source, and projectile, three kinds of gap become detectable with no apparatus beyond those traces—combinations never tried, concepts never pushed into their deepest tension, and ideas no concept in the knowledge base yet captures. On a demonstration corpus this surfaces concrete unexplored directions and, on the order of twenty, distinct nameable concepts absent from the 200-concept base. The map is not a diagnostic but an agenda: the same fracture that locates a gap steers generation into it. We show the map is generative—a detected blind cone, fed back through its own fracture, yields a real concept in the empty region (e.g. attention → “attentional commons”)—and a controlled within-concept test confirms it: targeting a detected cone steers significantly further into the void than the generator’s default fracture (p= 0.04), at no cost in judged value. Throughout, a two-level structure (source × fracture = theme; projectile = mechanism) organises the space, and we check honestly that a fracture steers the kind of idea, not its value.
No takes yet. Share an insight, caveat, or question.
Sebastian Wahl (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: