Human-evolutionary problem solving through gamification of a bin-packing problem

Abstract

Many complex real-world problems such as bin-packing are optimised using evolutionary computation (EC) techniques. Involving a human user during this process can avoid producing theoretically sound solutions that do not translate to the real world but slows down the process and introduces the problem of user fatigue. Gamification can alleviate user boredom, concentrate user attention, or make a complex problem easier to understand. This paper explores the use of gamification as a mechanism to extract problem-solving behaviour from human subjects through interaction with a gamified version of the bin-packing problem, with heuristics extracted by machine learning. The heuristics are then embedded into an evolutionary algorithm through the mutation operator to create a human-guided algorithm. Experimentation demonstrates that good human performers augment EA performance, but that poorer performers can be detrimental to it in certain circumstances. Overall, the introduction of human expertise is seen to benefit the algorithm.