Motivating Participants in Human-based Evolutionary Computation Systems.

Human-based evolutionary computation (humanbased EC), which uses humans as executors of all evolutionary operators, can solve problems for which only humans can judge the quality of solutions. In human-based EC systems, when participants whojoin problem solving are fixed, the upper limit of search performance of the human group should be unpredictable but roughly implicitly decided. Therefore, unlike standard EC using computers, human-based EC systems need to enhance or maintain motivations of participants for contributions to derive better search performance of the group. In the paper, we propose two methods for motivating participants in human-based EC systems. The first method is meant to enhance motivations by differentiating participants. More precisely, it feedbacks rankings on the number of times of producing and evaluating solution candidates to participants in a realtime manner. The second method is meant to maintain motivations by equalizing participants. More precisely, it sets the maximum allowed number of times of producing and evaluating solution candidates equally to all participants. It can also theoretically shorten a time period for problem solving. The two methods are strategically contrary to each other. We reveal though experiments that the first method enhances motivations of participants and also that systems using the second method and not using it produce same quality of solutions. These results suggest that we should feedback another rankings which are not based on the number of times to participants while using the second method to obtain high quality of solutions in a short period of time.