Conceptual Blending and Creativity
Creativity support tools aim to enhance the creativity of humans in a variety of activities. Such tools can range from simple objects (e.g., a chisel for enhancing creativity in sculpture) to more complex systems of representation (e.g., the Western musical notation system for enabling sophisticated music creation). During the last decades there has been a rapid emergence of software applications for facilitating creative processes in science, engineering or the arts. Examples include very popular platforms such as Google Docs for collaborative writing, Final Cut for video editing, Photoshop for image processing, SPSS for statistical analysis and so on. Naturally, depending on the task and the extent to which it is deemed creative or not, such systems can be instead called productivity support tools. For example, a typesetting system such as LaTex maybe be deemed a creativity support tool for writing a novel but merely a productivity support tool for putting together the proceedings of a scientific conference. In many cases, such systems aim to enhance both the creativity and the productivity of their users at the same time.
Productivity is much better defined and measured in contrast to creativity which is an abstract construct. Performance, time and error rates are some of the standard metrics used for productivity measurement (Shneiderman, 2007). At the same time, research on creativity evaluation has also identified a number of components (e.g., originality, domain competence, value, divergence, etc.) upon which creativity, be it human or computational, can be assessed (Jordanous and Keller, 2016). However, whereas scientific attempts to quantify and evaluate creativity have been ongoing for at least a couple of decades, the evaluation of creativity support tools is generally regarded a more recent field of research (Cherry and Latulipe, 2014). Usual approaches of study are the assessment of user experience utilising a framework of psychometric variables tailored for each task, the assessment of the artifacts that the users create through a panel of expert judges (Massetti, 1996) or both (Bonnardel and Zenasni, 2010).
This project attempts an evaluation of CHAMELEON's contribution to a melodic harmonisation task that belongs to the latter category. CHAMELEON (Kaliakatsos-Papakostas et al., 2017) is a melodic harmonisation assistant, that is a creativity support tool in the domain of music making. It is able to harmonise a given melody according to a number of different harmonic idioms and/or their blends. It does so by combining a generative implementation of the Conceptual Blending theory with statistical learning (Kaliakatsos-Papakostas et al., 2017). As a result, CHAMELEON, is capable of presenting a variety of diverse 'solutions' -some of which can also be rather unexpected- for melodic harmonisation at the push of a button.
The main hypothesis of this work is that this available diversity may potentially influence the perspective of a human user that performs a melodic harmonisation task. It is assumed that an on-demand stimulation of a user with a variety of a melodic harmonisation implementations may translate to a more explorative behavior from his/her side. To test this hypothesis, we designed an experiment where users performed a simple melodic harmonisation followed by a computationally supported melodic harmonisation on two similar melodies. The divergence of the produced harmonisations for each task was quantified using a number of idiom-independent harmonic similarity metrics. The user experience was also evaluated through post-task questionnaires that targeted to quantify the influence of a creativity support system in the manner of (Kantosalo and Riihiaho, 2019) and (Cherry and Latulipe, 2014). A prototype version of the software that supported these experiments can be found folling this link