TITLE:
Process-Oriented Understanding Estimation Using Code Puzzles
AUTHORS:
Hiroki Ito, Hiromitsu Shimakawa, Fumiko Harada
KEYWORDS:
Programming Education, Learning Analytics, Computational Thinking, Code Puzzle, Hidden Markov Model
JOURNAL NAME:
Creative Education,
Vol.13 No.3,
March
15,
2022
ABSTRACT: In the current programming
education, in order to assess the true ability of learners, instructors still have no choice but to monitor their answering
process, standing by them. However, this is impractical for freshman training
in educational institutions and newcomer training in companies. Because of the
practicality, a large number of learners are assessed at once using written
tests or Web tests. They usually inquire of learners whether they know
algorithms and grammar. If not, they assess only the behavior of source codes
they submit, at best. Under the training based on such assessment, in reality,
not a few learners fail to acquire the skill of writing source codes. It
implies that the attainment of programming skills cannot be assessed only by
tests on knowledge and submitted source codes. This paper proposes a method for analyzing learners’ understanding that
focuses on their thinking process of programming. The proposed method
focuses on a code puzzle in which learners arrange fragments of a program code
to satisfy given requirements. It aims to estimate the learner’s perspective on
how fragments are built up to achieve the requirements. Learners with low
understanding are assumed to be different from those with high in terms of the
consistency of arranging ways to compose code fragments for specific blocks in
source codes. For the discrimination, the method builds a model using a hidden
Markov model. The internal state obtained from this model would help
instructors grasp the learner’s understanding level. The results of an
experiment present that the hidden Markov model produces meaningful values,
which enable instructors to interpret the understanding of individual learners.