Abstract:

With the advent and population of Artificial Intelligence (AI) in education and ever-growing complexity of real-world tasks needed to solve in the 21st century, STEM education and practice require new visions and approaches to better understand the essential factors influencing the further advancement of this highly important educational paradigm. In this paper, we propose a meta-model for conceptual understanding the essence of the STEM domain.

Four basic concepts, i.e., STEM context, Agentic AI for STEM (human agent plus AI-based agent), STEM task and STEM process along with:
(i) models for each concept and
(ii) mutual interrelationships among the models are the components of the proposed meta-model.

Based on the thorough analysis of the literature, we motivate the role of the basic concepts, provide their definitions, and define their models. STEM task model is based on Wood’s (1986) and Efatmaneshnik and Handley (2018) works adapted to the STEM domain. Agentic AI for STEM model represents the human (student, teacher) - AI agent interaction when the task and process are conducted. The STEM outer context relies on global goals and known strategies. The STEM inner context model is expressed through extended TPACK framework to define technological, pedagogical, and content resources for STEM. STEM process model is an iterative sequence of cognitive and practical actions (Inquiry, design, experimentation, modelling, prototyping, analysis, reflection, etc.) used to solve authentic problems that integrate interdisciplinary knowledge for the 21st century. The essence of the approach is the interaction among the concepts’ models, where the STEM task is placed at the center. The following types of interactions are considered (bidirectional, unidirectional, cyclic, functional). Applying a computational vision to STEM (it focuses on Wing’s ideas and model transformations) as a research methodology, we have derived two generic patterns from the introduced meta-model, i.e., structural STEM pattern and behavior pattern. In terms of learning, both patterns that combine the structural and behavioral (characteristics of the four basic concepts are abstract generic structures to systemize to some degree and conceptually understand the domain. Furthermore, these patterns can be considered as an abstract teaching plan or scenario for use by a teacher in real STEM-driven learning environment. However, for doing so, patterns should be filled in with concrete content. The presented case study (Environmental temperature monitoring) empirically validates the proposed models and confirms the soundness and applicability of the approach.

Keywords:

STEM, Artificial Intelligence, Agentic AI, meta-model.