Testing Tournament Selection in Creative Problem Solving Using Crowds

Abstract

We tested a technique for creative problem solving, which used crowd-based genetic algorithms; one crowd generated initial ideas, another crowd evaluated the quality of these ideas, and yet another crowd combined pairs of ideas selected by the computer. The pairs were selected through a tournament method, in which two ideas chosen were biased toward higher quality. To test the technique, we asked a crowd to evaluate a subset of 468 solutions for the 2010 oil spill in the Gulf of Mexico produced by another crowd of 1853 individuals, and 311 solutions by 311 experts. The crowd evaluated the most creative crowd solutions as creative as the most creative expert solutions. Moreover, tournament selection led to greater improvement in the creativity of combined solutions than random selection, in which two solutions were chosen randomly. Creative problem solving using crowd-based genetic algorithms can work with good design.

Keywords: Creative problem solving, crowd-based genetic algorithms, evaluation