Abstract:

The recent development in Generative Artificial Intelligence (GenAI) environments, which use Large Language Models (LLMs) to mimic human interactions in chat-based interfaces, make it possible for even non AI experts to use AI, transforming the ways in which we search for and evaluate information. As a result, the ability to effectively use Generative AI (GenAI) tools is nowadays emerging as a new form of digital literacy. At the same time, GenAI environments emerge as promising tools in the field of education, enriching educational practice and increasing students’ motivation. However, little is known about how educators can integrate GenAI tools in educational activities in ways that promote the development of students' AI-related skills along with learning objectives. To address this issue in this study we employed a design based research methodology proposing the integration of a GenAI tool, ChatGPT, into collaborative concept mapping activities. In our design, three 12-year old students used ChatGPT as a co-worker in order to create dynamic conceptual models related to environmental education with Dynalearn, an intelligent learning environment that supports the creation of dynamic concept maps. By integrating GenAI into concept mapping activities we aimed to provide the students with a convergent context for evaluating and structuring new knowledge in a coherent and comprehensive way.

In the results of the study we describe how concept mapping supported two crucial aspects involved in the effective use of GenAI tool and often mentioned in the literature as challenging especially for primary students:
a) the formulation of purposeful prompts and questions, and
b) the critical processing and evaluation of the information provided by textual sources.

Along with these insights the study also aims to discuss guiding principles for the design of educational activities that might enable meaningful integration of GenAI tools while avoiding potential risks associated with their use.

Keywords:

Primary Education, ChatGPT, concept mapping.