In the last two decades, the development of areas such as Computer Networks and Artificial
Intelligence (AI) has favored the growth of other areas of knowledge, like Education. In this
area, new discoveries have changed the focus of research from old behaviorist educational theories
to constructivism, leading to a better understanding of how learning occurs. Meaningful
Learning (ML) is a constructivist theory in evidence nowadays and the Concept Map (CM) is
its main cognitive tool. Additionally, the recent developments on Distance Learning (DL) have
made it possible to apply the educational process in a larger scale.
In this thesis, automatic learning assessment mediated by concept maps is investigated. This
is related to a qualitative approach, named as formative assessment, which is compliant with
Bloom’s model, a reference for educational processes - teaching, learning, and learning assessment.
The proposal presented in this thesis is seen as an alternative solution to an important
issue in the area of Education: how to evaluate learning qualitatively, respecting each student’s
cognitive processes?
The integration of concept maps, domain ontologies, and genetic algorithms allows for advances
in automatic learning assessment and assistance. The paradigm of mere quantitative
assessment is broken, and a new approach to gradual and continuous assistance in learning is
presented. Following this approach, it is possible to accompany students individually, respecting
their idiosyncratic ways of learning, and also to group students based on specific cognitive
characteristics or development degrees.
This thesis begins a new research area, which can be synthesized as "Automatic qualitative
assessment of learning centered in Concept Maps, based on AI techniques: ontologies and
genetic algorithms". In this new research area, the thesis originated the following contributions:
² a prototype of an environment designed to aid teaching, learning, and learning assessment,
founded upon Meaningful Learning, encompassing a concept map editor, an ontology
editor, and an assessment module;
² A proposal concerning the use of genetic algorithms and ontologies in qualitative assessment/
assistance of learning, allowing for:
– step-by-step individual assistance;
– assistance to groups of students;
– comparisons among students.
Domain ontologies are generated by the teacher, who uses an ontology editor provided by
the environment. They comprise the structural knowledge that must be learned by students
before they can manage other forms of knowledge.
The genetic algorithm was designed to run in two distinct modes: i) generating multiple
CMs to compare with the student’s CM, allowing for learning assessment at any moment of
the course; this assessment is relative, centered in a determined number of concepts which
represent a partial structure of knowledge domain being studied.; and ii) generating an optimal
CM according to the ontology created by the teacher, to permit a complete assessment of the
learning of the knowledge domain which was studied. The proposed model was evaluated by
the implementation of prototypes for the assessment tool.
The genetic algorithm developed uses the ontologies as its search spaces. It emulates meaningful
learning cognitive processes, and constructs CMs that can be semantically compared to
that of the student. Its fitness function represents a way of measuring distances in the cognitive
field, being the measurement unit given by a taxonomy that organizes semantic dimensions
and, inside these, linking phrases. This taxonomy is used by teachers when they construct their
ontologies, and by students when they construct their concept maps.
The main challenges faced in the development of the research reported in this thesis were:
1) definition of a domain ontology model that could be applied to learning assessment; 2) definition
of a method and a scale that could be applied to the cognitive domain; and 3) definition
of a search mechanism in the ontology in accordance with constructivist theories of learning
assessment.
The research described in this thesis can be further developed with new functionalities or
improvements in functionalities already implemented. Some possibilities are suggested in the
end of the thesis, the main of which being the deployment of the environment in the Internet.
This thesis has generated 7 (seven) scientific contributions, 1 (one) in a qualis A magazine,
1 (one) in a qualis B magazine, 2 (two) in international congresses, and 3(three) in national
congresses. The results of this research advance what has already been attained by the
AmAm/UFPA research group, in whose context this thesis is inserted.
|
