Minimizing the Performance Gap: A Neural-Network-Inspired Error- Correction Model for Improving Student Achievement in Somali Higher Education

Abstract

Artificial neural networks improve their performance by repeatedly measuring the gap between what they produce and what is expected, then adjusting themselves to shrink that gap. This paper proposes that the same principle can be applied deliberately, and formally, to classroom teaching in Somali higher education. We define a simple performance-error formula for students, Error = Actual Marks − Desired Marks, and use it as the basis of a semester-bysemester feedback loop modeled on the forward pass, error computation, and backpropagation steps of a neural network. Rather than a metaphor alone, the paper sets out a concrete, low-cost procedure a lecturer can use each semester: set a desired target per student or course, measure the actual result, compute the error, prioritize interventions by the size and direction of the error, and re-measure in the following semester. We discuss how this model can be implemented with the limited infrastructure typical of many Somali universities, its expected benefits for both students and lecturers, and the practical limitations that must be managed. The paper is intended as a preview of a fuller empirical study to be carried out across upcoming semesters.