ISBN
:
978-602-97249-0-5
International Seminar on Vocational Education and Training
151
membership for an object is one, it means that the object is absolutely in that set. When the grade of membership is zero, it means that the object is absolutely not in that set.
Borderline cases are assigned to the values between zero and one. Precise membership grades do not convey any absolute significance. They are context-dependent and can be
subjectively assessed. Fuzzy set method has been applied in the assessment process. Echauz and Vachtsevanos 1995, presented a fuzzy grading method that utilizes
students’ and instructor’s performance measure to produce a ’fair’ mark distribution.
In a project based learning environment, students are encouraged to play an active role in evaluating their learning outcomes. Thus, the proposed assessment method
should also support to assess all performance. In formulating a grading system, it is necessary to recognize what considerations should be, directly or indirectly, part of the
grading policy. In this section, we explore class performance and ability and instructor’s technique Echauz and Vachtsevanos 1995:
Class’ Performance and Ability Instructor’s Technique
1. Absolute Performance: Each
student’s score is compared to a pre established model of what “good” or
“poor” performances signify. The fixed grade-boundaries method is based on
this type of performance only.
1. Teaching Effectiveness: The instructor
should lucidly expose the course subject. If he or she does a poor job of
guiding students through the learning process, how fair is it to compare their
grades with those of students who learned from a skillful teacher?
2. Relative
Performance: Each
student’s score is compared to others’ in the same class. The method of
equating top score to 100 is based on this type of performance only. All other
methods listed in Section I assign different weights to both absolute and
relative performances.
2. Tests’ Validity: A test is said to be
valid if it “tests” what the instructor really intends. One aspect of test
validity is balance of difficulty levels. A good way of designing well
balanced tests is to recognize Bloom’s
hierarchy of cognitive complexities ,
and include questions from the various categories. In general, valid tests can
be prepared by keeping track of discrimination ratios. These correlation
factors measure how effective each question is in being correctly answered
by students who really know the tested subject, and being incorrectly answered
by those who do not master it. Questions with high discrimination
ratios guide the instructor in the design of future test questions.
3. Score-Clustering: Gaps in the 3. Tests’ and Scoring Reliability: A test
ISBN
:
978-602-97249-0-5
International Seminar on Vocational Education and Training
152
score distribution often arise as a matter of chance, and do not necessarily
provide natural points of division
between discrete levels of ability.
However, given a scores’ histogram and a constrained range of mobility for the
grade-boundaries, it is reasonable for the instructor to visually place the
boundaries at points that separate clustered groups of scores, if such
points are within allowed ranges. Thus, local score-clustering provides a rule of
preponderant alternative in the grading process.
is said to be reliable if it consistently measures its goal. A student taking an
unreliable exam at two different times, is likely to score disparately. One
source of unreliability in tests is vagueness in questions, whence
multiple interpretations give rise to different answers. Another source of
unreliability arises in connection with the human scoring process. Are the
resulting scores reliable measures, or are they just random numbers? If the
question being scored is of the free response type, the instructor is likely to
be inconsistent by virtue of differences in mood and other human factors. Past
correlations between free-response and objective scoring can provide a basis
for determining how reliable the instructor’s subjective scoring is. To
this end, Spearman rank-order correlation coefficients are expedient.
4. Outside Performance: A group