Henry, JulieHernalesteen, AlysonCollard, Anne-Sophie2021-10-042021-10-0420212021http://dx.doi.org/10.1007/s13218-021-00733-7https://dl.gi.de/handle/20.500.12116/37481Although artificial intelligence (AI) is becoming increasingly important in the media environment (search engines, chatbots, home assistants, recommendation systems, etc.), the general audiences’ knowledge of it remains limited, which biases their representations. To compensate for this, some governments show an interest in teaching it from an early age. It appears that educational resources related to AI literacy in schools are most often focused on technical skills. However, the challenges of such education are also ethical and societal, requiring an interdisciplinary and critical approach. This research aims at developing a 10–14 years old curriculum questioning the concept of intelligence in AI systems, and crossing computer science education and media literacy education. Through a role-playing game, the children discover the basic concepts of machine learning. Beyond their initial representations, which they become aware that they are largely fueled by the media, they can realize that an AI system is the result of design choices and that it only works within the framework that has been defined for it. Moreover, the possibility for teachers to teach the curriculum themselves in their classes is also evaluated. To this end, the curriculum was taught to 60 future trainee teachers, 70 middle school pupils, and 12 elementary pupils. Interviews were conducted also with 5 teachers who had either observed the curriculum taught by a researcher or attempted to teach it themselves. The results show that the children’s representations have evolved towards representations that are more technically correct (although incomplete), but not very oriented towards aspects that open up critical questioning. The difficulties revealed in the implementation of the critical part are due in particular to the complexity of the IT concepts to be addressed, but also to the lack of teacher training. However, the data collected seems to confirm the interest and feasibility of crossing different disciplinary approaches to address certain aspects of AI. In conclusion, in addition to the curriculum, this paper describes a theoretical model of critical citizenship education in technology that integrates approaches to computer science education and media literacy education, and gives avenues for other designers and researchers to create AI critical educational experiences for K-12 learners.Artificial intelligence educationArtificial intelligence literacyComputer science educationConceptionCritical citizenship educationDesign-based researchInterdisciplinary approachMedia literacy educationPerceptionRepresentationText/Journal Article10.1007/s13218-021-00733-71610-1987