A Comparative Psychological Approach to Tiny-C Creativity

Creative intelligence has most broadly been defined by Currie and Halina (2019) as the ability to explore a hypothesis space for a given problem or domain and employ the resulting hypotheses to a novel situation. Such situations generally involve generating new knowledge from old - knowledge de novo. The products of such employments have been categorised by Kaufman and Beghetto's (2009) Four C Model into Tiny C, Little C, Pro C, and Big C creativity; this predicts a typology of behaviours from 'everyday' creativity like innovating a more efficient method to tidy one's desk through to major scientific and artistic breakthroughs like Darwin's theory of natural selection, or Gaudis Sagrada Familia. I shall be examining Tiny C Creativity using a methodology adapted from the Animal-AI Olympics. The AAIO is a competition being run this year by the Leverhulme Centre for the Future of Intelligence. Artificial Intelligence researchers submit algorithms to perform in a virtual world populated with tasks taken from the animal and child cognition literature. These tasks test the abilities of agents to perform 'creatively' in a novel situation in order to achieve some goal. The Aesop's Fable paradigm is one such task. This is a tool use task that requires the participant to drop items into water in order to raise its level so that they can obtain a floating reward (see Bird and Emery 2009; Taylor, Elliffe, Hunt, Emery, Clayton, and Gray 2011; Logan, Jelbert, Breen, Gray, and Taylor 2014, for use with corvids; see Cheke, Loissel, and Clayton 2012; Loissel, Cheke, and Clayton 2018; Clayton 2015, for use with children). This is presumed to be a novel situation for the participants and requires an understanding and an ability to predict the behaviours of physical objects such as stones and water. The AI algorithms are subjected to these previously unseen tasks and their performance statistically evaluated. I shall develop a computer game version of the virtual arena used in the Animal-AI Olympics. This will be done by programming in basic command functions. We may think of the 'gameification' of the arena as adding a trivial modality transducer, thus permitting us to directly compare performances by human users in the game with the performances of the artificial intelligence algorithms in the virtual world. The study will then involve children playing the game, receiving a reward incentive for successful completion of each task. Future studies will concurrently use psychometric measures for intrinsic motivation, fluid intelligence, and spatial cognition to enable a factor analysis of performance in the creativity game. Further studies will look at the developmental trajectory of Tiny C Creativity across age groups, to examine the predictions of German and Defeyter's (2000) suggestion that there is a U-shaped curve for creativity development, and to see if these predictions hold in the arena of Tiny C Creativity. These studies will be couched in a philosophical milieu, with the objective of informing contemporary AI research.