Divergent Association Testembedding edition

This is the Divergent Association Task from Olson and colleagues (PNAS, 2021), a validated measure of one slice of creativity: divergent thinking. The rules are simple and the scoring is pure geometry. Each of your words becomes a vector in latent space; your score is the average distance between every pair. Naming truly unrelated things is harder than it sounds, because your own associations keep pulling the words back together.

Following the original protocol, your score uses the first seven valid words; the extra three are insurance against invalid entries.

Pairwise semantic distance (higher and lighter = further apart). The diagonal is blank; a word is zero distance from itself.

Your score measures reach into open water, and reach is only the raw material. The full argument for where creativity actually lives, and the instruments to explore it, are here: The Shoreline Thesis. Your weakest link above is the best place to start reading from.

The official test at datcreativity.com scores with GloVe word vectors and human norms; this page uses a sentence-embedding model (all-MiniLM-L6-v2) running entirely in your browser, so the raw numbers are not comparable to published DAT scores. Instead the scale is self-calibrated: the low anchor is what you get by restating one idea seven ways, and the high anchor is a greedy search for the most mutually distant seven words in a common-English bank. One finding from the original research survives any embedding model: random words score frighteningly well, because randomness has no associations to fight. The interesting part of your score is that you produced it on purpose. Reference: Olson, Nahas, Chmoulevitch, Cropper & Webb (2021), "Naming unrelated words predicts creativity," PNAS 118(25).