Modelos de inteligencia artificial para asesorar el proceso evaluador de trabajos informáticos complejos

Abstract

We propose the use of artificial intelligence (AI) mo-dels to help evaluate complex projects in computerscience courses that involve technical and creativecomponents. The goal is to provide a methodology toanalyze the main variables that describe the evaluation,to discover possible biases and discrepancies, and togenerate appropriate rubrics that avoid them. The pro-posed methodology has been applied to a first-year in-troductory course on computer science, which is taughtin both computer science and mathematics degrees. Inthat course, the students must develop a web page according to some technical requirements (compatibilitywith standards, number of HTML and CSS files, etc.),but it also possesses a creative component (layout, ap-pearance, etc.). Optimized AI models have been de-veloped using evolutionary algorithms to identify themost important variables that took part in the evalua-tion of these projects during five courses. The resultsobtained allow us to draw conclusions about the eva-luation process, possible improvements in the impar-tiality of the evaluation, and the possibility of gene-rating suitable rubrics for grading assignments. Themethodology is applicable to projects in other sub-jects since this technical-creative typology is frequentin university tasks.