Evaluating computational creativity: a standardised procedure for evaluating creative systems and its application

This thesis proposes SPECS: a Standardised Procedure for Evaluating Creative Systems.

No methodology has been accepted as standard for evaluating the creativity of a system in the

field of computational creativity and the multi-faceted and subjective nature of creativity generates

substantial definitional issues. Evaluative practice has developed a general lack of rigour and systematicity,

hindering research progress.

SPECS is a standardised and systematic methodology for evaluating computational creativity. It

is flexible enough to be applied to a variety of different types of creative system and adaptable to

specific demands in different types of creativity. In the three-stage process of evaluation, researchers

are required to be specific about what creativity entails in the domain they work in and what standards

they test a system’s creativity by. To assist researchers, definitional issues are investigated and a set

of components representing aspects of creativity is presented, which was empirically derived using

computational linguistics analysis. These components are recommended for use within SPECS, being

offered as a general definition of creativity that can be customised to account for any specific priorities

for creativity in a given domain.

SPECS is applied in a case study for detailed comparisons of the creativity of three musical improvisation

systems, identifying which systems are more creative than others and why. In a second

case study, SPECS is used to capture initial impressions on the creativity of systems presented at a

2011 computational creativity research event. Five systems performing different creative tasks are

compared and contrasted.

These case studies exemplify the valuable information that can be obtained on a system’s strengths

and weaknesses. SPECS gives researchers vital feedback for improving their systems’ creativity,

informing further progress in computational creativity research.