Solving the global STEM educational crisis using Cognitive Load Optimization and Artificial Intelligence–A preliminary comparative analysis

Abstract

There is a persistent STEM educational crisis exemplified by low student enrolments, and both high failure and attrition rates. ChatGPT is easy to use, however pedagogical quality is not necessarily assured. In one experiment the output had a high cognitive load exacerbated by cognitive gaps making the material hard to teach and learn. ChatGPT is a useful pedagogical technology but not a learning theory. Science, technology and engineering all start by quantitatively modelling systems in order to make accurate and quantitative predictions prior to construction or system modification. By contrast, the current learning theories in use today are based on qualitative soft-science principles, with subjective guidelines that are open to interpretation, which can lead to wide variations in the quality of instructional materials and learning outcomes. Cognitive Load Optimization (CLO) is a new Science of Learning (SoL) theory that quantitatively models relational knowledge as coherent, contiguous, pedagogically scalable schemas optimized for the lowest cognitive load. CLO schemas represent the easiest, fastest and most efficient learning paths and are the fundamental basis of instructional design and teaching. Because CLO schemas are pedagogically scalable it is possible to create CLO schemas that are contiguous across different educational levels (school, college and university) thereby uniquely meeting the goals of the American National Science Foundation SoL (‘optimized learning for all’) and the Australian Grattan Institute (‘optimized learning from pre-school to university’). Using CLO results in significant improvements in STEM learning outcomes but is a detailed methodology that can be time consuming to use. The relative advantages and disadvantages of ChatGPT and CLO are highlighted.