Abstract of my research to date - submitted to the London Hopper Colloquium in May 2009

Evaluating Computational Creativity: Excuse me, Computer, How Creative Are You?

Anna Jordanous

DPhil researcher

Music Informatics Research Centre / Creative Systems Lab

Department of Informatics

University of Sussex

a.k.jordanous@sussex.ac.uk

Creativity is a highly subjective concept; people have an intuitive understanding of what creativity is, but may struggle to capture what the word ‘creative’ means in a formal definition. This poses problems should one want to describe how creative something is, or compare it against another item in terms of the creativity being demonstrated. Such descriptions and comparisons are vitally useful both for reflections on previous creative work, and comparison of progress in creative-led practice. In particular, when looking at examples of computational creative systems, how can one machine-produced artefact be compared against others? Or against human-produced artefacts? To measure creativity I use several related aspects of creativity, such as novelty, usefulness, fluency of ideas etc., which are more tightly defined and therefore easier to evaluate (either by assessing people’s opinions or by using quantifiable tests). My research offers a set of tools to measure creativity from a variety of viewpoints: from statistical tests of how unique a creative piece of work is in a given domain, to social opinions on the significance and usefulness of that piece of work. These tools form a methodology for a standardised evaluation of creativity across many domains, for comparison and reflection.