Contribution: This paper presents the "PBL classroom model," an agent-based simulation (ABS) that allows testing of several scenarios of a project-based learning (PBL) application by considering different levels of soft-skills, and students' perception of the methodology. Background: While the community has made great advances in conducting engineering education studies, achieving goals of meaningful learning is still difficult. In this context, each new teaching methodology should be field tested, which can be time consuming. Introducing a simulation that can pre-test methodologies to check their efficiency can assist in the methodological building process.Research questions: The proposed simulation environment helps in understanding the differences in the learning rate of a group of agents (students) with different levels of humanization and instruction. Methodology: Partial least squares path modeling (PLS-PM) was used to build the causal relationship between the students and a PBL methodology, to generate the input data for the simulation. Based on the analysis of the responses to a survey performed over five consecutive semesters, it was possible to generate some cause-effect data parameters to be used as the input for a hybrid type of ABS, known as PLS agent.Findings: The use of the proposed model, based on a computational approach, to study the educational process was validated; the simulation outcomes indicate that higher levels of self-esteem, self-realization, and cooperation result in better learning experiences.