A Network-Based Computational Model for Creative Knowledge Discovery Bridging Human-Computer Interaction and Data Mining

Abstract

As a branch of computational creativity, Creative Knowledge Discovery (CKD) aims to search for valuable, previously unknown, or ignored, relationships between concepts, and create new patterns by taking advantage of existing patterns or by analogy to patterns in other domains. Data mining has been widely used in CKD research. However, most proposed mining algorithms lack a theoretical basis for computational creativity due to their origins in traditional knowledge discovery in databases (KDD), which stymies novelty. In addition, integration of human-computer interaction (HCI) is often overlooked for assisting discovery of creative knowledge despite the human end user possessing problem solving intelligence. To address these issues, a network-based computational model bridging human-computer interaction and data mining is proposed arising from an initial investigation on the theoretical basis of computational creativity. A corresponding creativity evaluation methodology, Multi-dimensional In-depth Long-term Case studies (MILCs) is also introduced. In order to evaluate the proposed model, a web tool called B-Link has been developed. Longitudinal interviews and a questionnaire survey have been conducted by applying the MILCs evaluation method. The success of finding novel items and obtaining inspiration in interviews as well as the positive survey rating results of all five creativity metrics have suggested that B-Link is able to guide thinking processes and aid creative knowledge discovery effectively, which demonstrates the capability of the proposed network-based computational creativity model integrating human-computer interaction and data mining.