---

TALK: “Measuring success in open ended-learning AI models: what has been learned from computational creativity evaluation research”

Abstract
Evaluation of success is vital for scientific progress. Success in open ended learning is difficult to specify or measure in an interpretable or psychologically valid way, and progress is driven by goals other than achieving benchmarks or intended targets. There is quite a natural comparison to how evaluating creativity presents the same challenges. I will discuss how computational creativity research handles evaluation of creative software, including the practical challenges that are not faced in other types of AI evaluation. While Computational Creativity and creative AI researchers have had some success in agreeing evaluation approaches, there are also many unsolved issues that remain unsolved. Many past methods fail to scale up to current generative AI. Yet all is not lost; I will finish by reflecting on what I find exciting for the future in such evaluation.

---

Short Bio

Dr Anna Jordanous is a Reader and Deputy Head of School, in the School of Computing at the University of Kent. She is a member of the Artificial Intelligence and Data Analytics (AIDA) research group. Her research areas include computational creativity and its evaluation, music informatics, digital humanities, knowledge modelling, Semantic Web, and natural language processing. Primarily she works with computational creativity - the modelling, simulation or replication of creative activities and behaviour using computational means - with a focus on the question of how to evaluate claims of computer software being creative. As well as writing creative software to improvise music, Dr Jordanous has contributed a highly-cited standardised procedure for evaluating creative systems. She also uses music information retrieval and natural language processing in her work.