Generative AI in linear programming teaching: Errors, learning, and collective construction

Abstract

This paper presents an innovative pedagogical experience in teaching linear programming model formulation, incorporating Generative Artificial Intelligence (GAI) as a support tool in a university context. The proposal was carried out in the Quantitative Methods for Decision Making course, at the Faculty of Economic Sciences of the National University of Córdoba, Argentina, within the framework of an active learning approach and using the Problem-Based Learning (PBL) methodology. A total of 98 students participated, organized into groups, and interacted with ChatGPT to address tasks related to the mathematical modeling of complex problems, critically analyzing, correcting, and validating the ansewrs obtained. The objectives of the experience focused on three axes: to strengthen the understanding of the principles of mathematical modeling and linear programming; to develop a critical attitude toward the solutions proposed by GAI; and to foster collaborative work and the responsible use of technology in academic and professional contexts. The activity was structured around group problem-solving, in which teams had to submit their queries to GAI, examine the proposed formulations, and make any necessary adjustments until they achieved a correct model. The process included identifying errors, analyzing the objective function and constraints, and justifying each decision. All interactions with the tool were recorded and analyzed, and the final product consisted of a group report on the experience carried out. A qualitative analysis was conducted using group surveys with open-ended questions, whose responses were coded and organized into four categories: practical usefulness, quality of responses, ease of use and learning, and teamwork. Among the most relevant findings, it was highlighted that ChatGPT facilitated the formulation of models and the organization of information. However, its effectiveness was limited in specific situations, where some errors arose during the mathematical modeling stages. Regarding learning outcomes, the experience showed that GAI can be a valuable resource to enrich the understanding of linear programming and promote the development of critical thinking, provided that students have a solid theoretical foundation that allows them to validate and correct the responses. Regarding collaborative work, the group dynamics fostered discussion of ideas, exchange of perspectives, and collective knowledge construction, all valuable aspects in massive courses where individual activities tend to predominate. It is concluded that the experience demonstrated the potential of GAI as a support in teaching-learning processes related to linear programming, also emphasizing the importance of its critical and responsible pedagogical integration. Although the tool offers speed and accessibility, its effectiveness depends on the user’s prior knowledge and constant human validation. This practice aligns with the Sustainable Development Goals, particularly SDG 4 on quality education, and contributes to the ongoing reflection on the role of emerging technologies in higher education.