2026 ASEE Annual Conference & Exposition

Responding to the growing call to integrate artificial intelligence (AI) into undergraduate curricula this course module was designed to introduce students to the foundational principles underlying modern AI systems. The goal was to help students develop a conceptual and operational understanding of the agents they encounter on an increasingly frequent basis in their engineering careers and everyday lives alike, while also recognizing the sources of error and bias that affect model outputs. The existing Biomedical Computing course, which culminates with study of numerical solutions for partial differential equations, was identified as a natural setting for this integration.
The neural network module takes place over six class periods and three graded assignments and is designed towards the outcome ‘students will understand the core concepts behind generative AI agents and evaluate the limitations and sources of error within these models.’ The first session establishes a shared conceptual foundation through an instructor-led lecture introducing neural network structure, terminology, and the conceptual parallels between biological and artificial learning systems. Students also interact directly with several generative AI models to observe their behavior and limitations. The second session builds on this foundation by guiding students through neural network training and data preparation using MATLAB’s built-in machine learning toolbox. Following this class, students complete an individual homework assignment in which they train a neural network, providing an accessible visualization of how input data are weighted and refined through iteration. Subsequent sessions focus on a two-part project that transitions students from using preconfigured tools to constructing and analyzing their own models. In the first part, students normalize data for input into a provided MATLAB neural network, train the model, and evaluate its accuracy and sources of error and provide a brief executive summary outlining their work. After demonstrating mastery of these skills students progress to the second part of the project in which they design, train, and evaluate a custom multi-layer neural network and then compare its predictive performance to that of the MATLAB-generated model. At the completion of part two, students provide prediction results, quantitative error analysis, and a written report describing model constraints, design decisions, justification for those choices, and proposed next steps for improvement.
This activity illustrates an adaptable approach to integrating AI literacy into a computationally focused biomedical engineering course without displacing essential disciplinary content. Successful implementation depends on access to computational tools, instructor familiarity with AI modeling, and student readiness in numerical methods. Nevertheless, the framework can be readily transferred to other computing, data science, or engineering contexts to help students critically engage with both the capabilities and limitations of artificial intelligence.