Crowdsourcing approximates convex optimization
If you use the many small decisions made by a crowd to inform a search process through a latent space, you can approximate convex optimization. Optimizer-specific dynamics such as momentum can also be integrated to determine user prompts for optimal convergence. In the case of a creative task, this is consonant with the view of creativity as search. That said, crowdsourcing might also be useful for open-ended tasks.