Open and closed-ended problem solving in humans and AI: The influence of question asking complexity

Abstract

Question-asking, an underexplored aspect of creativity, is integral to creative problem-solving and information-seeking. Previous research reveals that lower creativity correlates with asking simpler, closed questions, while higher creativity correlates with complex, open-ended inquiries. The present study explores the relation between question asking complexity and problem-solving tasks involving open- and close-ended thinking and how these abilities generalize and compare to AI. In Study 1, participants (N = 89) completed the alternative questions task (AQT), a close-ended riddles task (Stumpers), and the alternate uses task (AUT), a creativity measure. Our results show AQT question complexity wasn't correlated with stumpers performance, although it correlated with AUT originality (r = .3). In Study 2, participants (N = 100) completed the AQT, AUT, and open-ended creative problem-solving (CPS) task. CPS responses were evaluated for originality and quality. A positive correlation was observed between CPS quality and AQT complexity (r = .29) and originality (r = .34). In study 3, AI agents (N = 100) completed the AQT, AUT, stumpers, and CPS tasks. Like humans, AI's AQT originality and complexity were related with open, but not closed problem-solving. AI questions were also significantly more creative and complex, it solved more stumpers and gave higher quality CPS solutions. Surprisingly, human and AI CPS originality didn't differ. We find significant links between question complexity and open—but not closed-ended—problem-solving in humans, which generalize to AI. Our results highlight the significance of complex and creative question-asking in everyday life and as an integral part of our problem-solving toolkit.