Thinking with things that learn

This paper investigates the relationship between thinking and the human interaction with things which, in the last decades, was greatly enriched by the diffusion of computational technologies. The analysis of this relationship is reported with a focus on the type of things that might be involved in the interaction, from daily life objects and materials, to computational artefacts, and to computational artefacts able to learn. In particular, machine-learning is presented as an emerging design material able to enhance thinking by fostering a reflection on how machines can learn, on their identity and on the qualities of the input from which they learn. These considerations are at the basis of an exploration of machine learning as a design material for the development of learning artefacts for children. This investigation was carried out by adopting a Research through Design approach, particularly characterized by practice-based design activities. These consisted mainly in prototyping, from low-fidelity paper mock-ups, to physical computing prototypes, to playing with an open-source machine learning software. This process resulted in the development of two artefacts, Shybo and Pinocchietto, that were used as part of two different playful learning experiences with children in primary schools. The two activities were characterized by a different use of the involved robot. In the first case, Shybo supported a reflection on colours and sounds. In the second case, Pinocchietto was used to reflect on the similarities and differences between machines and humans, and to reflect on the robot functioning by formulating hypothesis and testing them out. The two artefacts and the related experiences are reported with the aim of contributing to the understanding of the learning ability, machine-learning in this specific case, as a design material to support thinking. To this end, the final part of the article reports observations regarding the activities providing insights about how the artefacts were perceived, children attitude toward the experience, and machine learning features. These observations are also used to introduce also emerging design opportunities.