Psychology of Assessment - Week 14
Computerized Testing and Test Bias
Computer-Assisted Assessment
Basic Issues
- Includes testing that is computer
administered, scored or interpreted
- Computer Adaptive Testing - the computer
decides what portion of a test gets administered, called item
branching
Advantages of Computerized Testing
(from Butcher, 1987)
- Economy of professional time
- Potential to have trained individuals
oversee test administration
- Administration, scoring and interpretation
can occur quickly
- Elimination of scoring errors
- Interpretations can be
standardized
More Advantages of Computerized
Testing
- Ability to build and access large data
bases
- Can be adapted to special
populations
- Less distraction (can view only one item at
a time)
- Decreased test administration
time
Disadvantages of Computerized Testing
(from Butcher, 1987)
- May intimidate the test taker
- Generally must answer an item before you
can move on
- Answering missed items may not be
allowed
- Long learning curve for the
examiner
More Disadvantages of Computerized
Testing
- Must invest in
hardware/software
- Reduced opportunity to observe the
examinee
- Lack of validity supporting the use of
interpretive programs
Computer Input
affected by...
- Item type and item content
- Ability to skip items
Computer Output
considerations...
- Is the computer printout used as the
report?
- What type of language is used?
- Did the computer interpret the
test?
Legal, Ethical & Professional
Issues
- APA requires that a trained clinician make
behavioral observations
- Untrained professionals should not have
access to these tests
- Psychological tests are controlled by
states, but related tests are not
- Computer error can lead to
misdiagnosis
- Confidentiality must be upheld
Test Bias
- Adverse Impact - tests must not
systematically reject minority applicant in job
screening
Adverse Impact
affected by
- Item Content - no longer thought to be the
culprit
- Differential Item Functioning Analysis -
Find and eliminate items that discriminate between
minorities/nonminorities
Avoiding Adverse Impact
- Compare regression lines - If the slopes
are equivalent then the test is not biased
Measuring Adverse Impact
- The 4/5ths rule - Selection rate of any
race, sex, ethnicity less that 80% of the rate for the group with
the highest selection is evidence of adverse impact.
New Yorks Truth in Testing Law
Test publishers must...
- Disclose studies on validity
- Tell test takers what their scores
mean
- Provide test score and correct answers upon
request
Back
to the Psych of Assessment Course Outline