Are Two Heads Always Better Than One? Human-AI Complementarity in Multi-criteria Order Planning

Innovative solutions are often crafted through synergizing contributions from a group of diverse entities with complementary strengths and weaknesses. The same is expected to persist with the advent of artificial intelligence (AI). Thus, this work aims to investigate whether the synergistic interaction of human decision makers and optimization (AI) algorithms can significantly improve the solving of challenging, multi-criteria order planning problems. To this end, a Human-AI complementarity framework leveraging on emerging transfer optimization methods is first put forward, enabling the adaptive reuse of experiential priors to inform search. Next, empirical analysis on a carefully designed multi-criteria order planning problem is conducted. Finally, 3 key insights arising from situations where the human supplied prior is perfect, imperfect, or flawed are discussed to adequately address the research question posed.