Should we pursue SOTA in Computational Creativity?

Should we pursue a state-of-the-art in Computational Creativity? The activity of ‘SOTA-chasing’, or working towards beating performance standards achieved by the current state of the art, is typical in many research disciplines relevant to computational creativity such as Machine Learning or Natural Language Generation (SOTA). Computational Creativity (CC) research does not typically engage with SOTA-type benchmarks. Consequently, it becomes harder to objectively identify high-performing systems in a creative domain (area of creative application), despite our research efforts building significant bodies of work in several domains. This paper critically engages with the use of SOTA in other related disciplines and explores the idea of working with SOTA-based evaluation in CC. The paper offers recommendations for (careful) use of SOTA to invigorate and direct CC progress.