Page 30 - 2023-Vol19-Issue2

26 |
also been mentioned in [7] where it is necessary to compare
one process model with other process models. Furthermore,
to make comparisons, an approach is used using the process
cube [8]. Thus, each dimension can be compared to determine
whether there is discrimination or not in the assessment of
lectures.
II. RELATED WORKS
Research on fairness learning analytics is increasingly at-
tracting the attention of researchers who are concerned about
education. Studies related to fairness to answer some unfair
problems to process modeling that can lead to unfair and
detrimental result. Related to educational research of fair-
ness, there are some discoveries. Based on [9], the results
indicate that the teachers’ conceptions of fairness were in-
fluenced by three reasons: (a) individual mechanisms, (b)
social mechanisms, and (c) dialectical relationships between
individual and social mechanisms. Each of them was the fac-
tor from teacher’s perspective that influenced the assessment.
Moreover, the study [10] yielded three categories: distributive
justice, procedural justice, and interactional justice. The study ML-based decision systems. This study provides an open-
concludes by proposing a range of implications for different
testing stakeholders.
Mahnaz Sadat Qafari and Wil van der Aalst in the Fairness-
Aware Process Mining research, the results of the application
of machine learning and data mining techniques are not always
acceptable. In many situations, this approach tends to make an
overt or unfair diagnosis and applying it can lead to erroneous
or even discriminatory conclusions. This study presents a
solution to this problem by creating a fair classifier for such
situations. Unwanted effects are eliminated at the expense of
reducing the accuracy of the resulting classifier [2].
Research by Qian Hu and Huzefa Rangwala highlights
concerns about model bias and injustice causing unfair out-
comes for some groups of students and negatively impacting
their learning. This study demonstrates by concrete examples
that biased educational data leads to biased student modeling.
This research encourages the development of equitable for-
malization and algorithms for educational applications. The
proposed formalization is individual fairness and group fair-
ness. This study proposes a model based on the idea that
predictions for students (identifying students at risk) should
not be influenced by their sensitive attributes. The proposed
model is proven to be effective in removing the bias from this
prediction and hence, making it useful in helping all students
[1].
Other research examines possible forms of discrimination,
as well as ways to measure and define fairness in Virtual
Learning Environments (VLE). Prediction of student course
outcomes was carried out on the VLE dataset and analyzed
with due regard to fairness. Two measures are recommended
Andreswari & Syahputra
for investigation of previous learning data, to ensure their
balance and suitability for further data analysis [11].
Subsequent research proposes a systematic process for
framing, detecting, documenting, and reporting the risk of
unfairness. The results of a systematic approach are combined
into a framework called FairEd, which will help decision
makers to understand the risks of unfairness along the environ-
mental and analytical fairness dimensions. This tool makes
it possible to identify data containing a risk of unfairness,
identifying models to see that potentially unfair outcomes can
be reduced without compromising performance [12].
Furthermore, the following research advocates the use of
simulation as the main tool in studying algorithm fairness.
The study explores three examples of previously studied dy-
namic systems in the context of equitable decision making
for bank loans, college admissions, and attention allocation.
By analyzing how learning agents interact with these systems
in simulations, it is shown that static analysis does not pro-
vide a complete picture of the long-term consequences of
source software framework that can be extended to implement
fairness-focused simulation studies and further reproducible
research [13].
III. METHODS
In the process of assessing the website application develop-
ment course, there are several factors that are considered to
assess fairness in determining student graduation. These fac-
tors include student attendance, lecturers, gender, and student
batch and assessment index. Online lectures result in students
and lecturers not being able to meet in person, but attendance
can still be calculated through activities carried out during lec-
tures. The existence of differences in lecturers in each parallel
class is also possible to have an impact on assessment bias
due to different standards. Although using the same semester
learning plan, the implementation and treatment of each lec-
turer may be different. Furthermore, gender differences are a
factor that is often used as an identification of discrimination
in an assessment [12]. The last factor to be considered in
this study is the class, where students who repeat or are more
senior are also possible to be one of the causes of differences
in assessment standards. Assessment in the form of numbers
is not included because it is hypothesized that the assessment
index is carried out in a discriminatory way without looking
at the scoring rubric. Based on Fig. 1, the causal factors of the
lecture process discrimination in a fishbone diagram can be
divided into machine, man, measurement, method and milieu.
Material is not included because the lecture process has no
material differences and moreover the data is not recorded in
the lecture.