Is Explainability a Prerequisite for Creativity?

抄録

In this paper we hypothesise that providing an 'explanation' about a piece of programmatically generated art can enhance its social value and perceived creativity. For Machine Learning generated art, this could mean explainability may be a key component in whether it is considered to be creative. Here we describe a pilot study testing this hypothesis through the evaluation of generated musical pieces. 29 people participated by completing an online survey. Participants were divided into two groups, one of which was given a fabricated 'explanation' for each piece. There were statistically significant differences in each group's perceptions of the generated pieces, which lends support to our hypothesis. However, there was no significant difference between each groups' answers when asked how creative they thought the program that generated the music was.