Artificial intelligence tools for student learning assessment in professional schools

By Almeida, P.; Novais, P.; Costa, E.; Rodrigues, M.; Neves, J.



The necessity to maximize the learning success of the students as well as to produce professionals with the right skills to fulfil the market requirements, raises the question of closely following and assessing the learning paths of the students of Professional Schools. To solve at once problems and difficulties that arise during the learning process, we need to develop technologies and tools that allow the monitoring of those paths, if not in real time, at least periodically. Supported on a knowledge base of student features, also called a Student Model, a Student Assessment System must be able to produce diagnosis of student’s learning paths. Given the wide range of students’ learning experiences and behaviours, which implies a wide range of points and values in students’ models, such a tool should have some sort of intelligence. Moreover, that tool must rely on a formal methodology for problem solving to estimate a measure of the quality-ofinformation that branches out from students’ profiles, before trying to diagnose their learning problems. Indeed, this paper presents an approach to design a Diagnosis Module for a Student Assessment System, which is, in fact, a reasoner, in the sense that, presented with a new problem description (a student outline) it produces a solved problem, i.e., a diagnostic of the student learning state. We undertook the problem by selecting the attributes that are meaningful to produce a diagnosis, i.e., biographical, social, economical and cultural data, as well as skills so far achieved, which may drive, as constraints or invariants, the acquisition of new knowledge. Next, we selected the metrics that would allow us to infer the quality of the ongoing learning, i.e., the degree of expertise on the currently attended learning domains. To collect these indicators we used the Moodle e-Learning System. Both, attributes and metrics, make the student model. Finally, we designed a reasoner based on Artificial Intelligence techniques that rely on the Quality-of-Information quantification valuations to foster a Multi-Valued Extended Logic Programming language, a key element in order to produce diagnosis of the student learning paths. Confronted with a new case, i.e., a student model, the reasoner evaluates it in terms of its QI and outputs a diagnostic.


Google Scholar: