Adaptive Testing

Someone on a Facebook group recently posted a video from McGraw Hill about "adaptive testing". McGraw Hill is the company writing SBAC, which is the Common Core test being given here in Connecticut.



And the conversation that followed went like this:



It's important to remember that while the ability to compare student/student school/school might be something that other states hope to accomplish, here in Connecticut that is not the stated goal of the testing.

Here in CT the goal of the testing is to ascertain "mastery":

Sec. 10-14n. Mastery examination. (a) As used in this section, “mastery examination” means an examination or examinations, approved by the State Board of Education, that measure essential and grade-appropriate skills in reading, writing, mathematics or science.

So in CT the words "standardized testing" are irrelevant, since a "mastery examination" can theoretically be an adaptive one:

Reading passages might vary from student to student according to the complexity of the language and concepts they contain. But all test-takers would be asked to engage in grade-appropriate higher-order thinking skills, such as making inferences about what characters in the passage might do next.

It's important that computer-adaptive testing be standards-based, meaning there is a blueprint that ensures every kid sees test questions that reflect the full breadth and depth of on-grade-level content.

Interestingly though, Pearson Publishing, the company responsible for writing PARCC, the Common Core test being given in states that are not using SBAC as we do here in Connecticut, writes that:

By definition, adaptive tests given to different individuals vary in their statistical specifications and, as such, cannot be considered equivalent in the strictest sense.

...we encourage the recognition that adaptive testing in the context of the common core assessments must be pursued cautiously and deliberately.

There are trade-offs with choosing to use adaptive testing over conventional testing. Adapting testing... does not produce tests that are equated in the strictest statistical sense. It is important to understand the threats to the validity of scores and interpretations resulting from adaptive testing...

So if I'm understanding this correctly, the company competing with SBAC's McGraw Hill, is telling us that adaptive testing may not be the best way to test for mastery, since each test is different for each student.

This all lends a lot of validity to a comment that another person left on that Facebook post:




Also pointing out that adaptive tests have more of a chance of being invalid is a parent, as quoted by Diane Ravitch on her blog, who states:

My kid doesn’t like online adaptive assessments. She likes knowing there are 50 questions in 40 minutes.  She hates tests that give you many more difficult questions when you answer correctly. The test seems to go on forever.

So, one time she decided to hit buttons randomly and get a bunch wrong. Then the computer spit out fewer, easier questions, and she was able to finish the test at last.

Stupid politicians and administrators seem to forget that kids are intelligent, and when they know the tests don't count towards their grades they don't try their best; they just figure out ways to get through it all faster. One teacher in Connecticut actually admitted to me that she subtly suggested that her kids randomly answer things just to get through it (of course at this point the testing is not connected to teacher evaluations in CT, and once that happens she will likely stop advising her kids to do that).

But the whole idea of testing the kids for mastery, or whatever you want to call it, is a stupid one. The bottom line is that the kids know it doesn't count toward their grade, so they don't take it as seriously. And tests don't accurately measure the academic capabilities of kids the way they are supposedly meant to:



So the best thing a parent should do is to continue to refuse to let your child take them.