Today I found a quick creativity test, called the Divergent Association Test (paper - https://www.pnas.org/content/118/25/e2022340118) It's pretty neat and only takes a couple minutes to take. You are tasked with typing a series of nouns that are as unrelated to one another as possible. The idea, I imagine, is that creative people see associations others do not, and therefore also can see how a series of things aren't associated with one another.
As I guessed, word embeddings power the semantic distance calculation between the words people enter, more precisely it uses GloVe:
Objectively measuring the relatedness of words is difficult, so as a proxy we look at how often the words are used together in similar contexts. We use the Common Crawl corpus, which contains thousands of different words across billions of webpages. Using [a word embedding] algorithm, we compute the distance (or relatedness) between the words; words such as “cat” and “dog” are often used close together and thus have smaller distances between them, while words such as “cat” and “book” would have greater distances. The total score is simply the average of these word distances: greater distances give a higher score.
This made me wonder if an even more constrained version of the 'test' could also be useful, like name the types of activities you like to do that are as different from one another as possible, as a way to generate ideas for learning how to do very different things.