+
Related Flashcards
Related Topics
Cards In This Set
| Front | Back |
|
A test is biased if there are subgroup differences in test score (eg. race) that are unrelated to the constuct being measured
|
Test biases
|
|
__ would occur when test scores are different for certain subgroups (eg. gender), and these differences are not related to differences in actual ability levels
|
Bias
|
|
What are the 3 issues of test biases?
|
–It’s the meaning inferred from test scores that makes a test biased, not the test itself.
–Bias is normally dealt with at the aggregate or group level – not on an individual case-by-case basis.
–Bias occurs when groups have same underlying scores on the construct, but test scores erroneously imply the groups are different.
•If groups had different underlying scores, and the test accurately conveyed this, the test is doing its job.
|
|
Is using a test that detects true group differences bias?
|
No but it does incolve ethical, political and philosophical issues (ie.fairness issues)
|
|
Test fairness breaks down into 4 components
|
- test bias
- social/political issues & values - probability of success does not = probability of being hired - differences in predictor-criterion relationships by subgroup |
|
What is the third facet of test fairness?
|
•Test is unfair if people of equal probability of success on a job have an unequal chance of being hired
|
|
What is the fourth facet of test fairness?
|
-single-group validity
|
|
–Test predicts for one group but not another
|
Single -group validity
|
|
–Test predicts for both groups but better for one
|
Differential validity
|
|
A test would be unfair if... x3
|
–Test bias was shown (difficult to show)
–Single group validity (rarely occur)
–Differential validity (rarely occur)
|
|
–Intercepts or slopes are different
|
Differential prediction
|
|
Is a differential prediction on a test unfair? Afterall the regression lines demonstrate the tests were both criterion valid?
|
No because we may still have adverse impact
|
|
Can you avoid adverse impact by adding a constant to a subgroup’s scores?
|
No, that is illegal
|
|
Is adverse impact in the case of differential prediction fair in a strict scientific standard?
|
Yes
|
|
Is adverse impact in the case of differential prediction fair from a social or ethical perspective?
|
No
|
