What a New Test Needs to Have

I’ve prepared millions of students for the SAT and ACT with classes, books and software. Those tests built my house. Yet I’ve been calling for the elimination of the SAT and ACT for decades. Now, with the University of California’s decision to eliminate their use over the next four years, it seems the reign of the standardized tests is ending.

But a world of college admissions without some common yardstick is problematic, too. Any number of experiments have shown that without some form of standardized testing, grade inflation will be immediate and severe, since high schools want to help improve their students’ college prospects. Efforts of parents, teachers and administrators will render the only remaining “objective” admissions criterion meaningless. Further, colleges will be forced to give more weight to application essays, which are often of dubious authorship.

So what should replace the SAT and ACT? As someone who has studied these tests at a molecular level -- and might have thought about them more than anyone alive -- I have a few ideas.

First, we need tests that avoid the largest problem underlying the SAT and ACT: the presumption that there is a small body of knowledge that every student should acquire before beginning college. That, in the case of the SAT, led to 96 years of standardization -- not of the tests but of students. Kids have different interests; forcing them all to learn the same thing sucks the joy, and much of the effectiveness, out of K-12 instruction. Further, the people who decide what every student should learn are the same folks who gave us new math, bilingual education, whole language and the Common Core. Not a single national curricular effort has met its goals in any measurable way.

Some tests, trying to avoid the forced march of what is essentially a common curriculum, measure what they imagine to be “basic competency” in a subject, but that’s even worse. Schools and students, understanding that the only skills that matter are the basic ones, teach those to the exclusion of everything else. High school will become a second junior high school.

What colleges need is a system that encourages excellence without mindless standardization, by allowing students to be tested in a wide variety of subjects, even in different approaches to each subject -- for example, a computational math approach versus theoretical one.

Many psychometricians argue that tests should be measured primarily by accuracy, precision or predictive validity. More important than any of these, high-stakes tests should be judged by the kind of education they encourage. Students will prepare for any high-stakes test, but what will their preparation look like? With the right sort of tests, preparation will look like the acquisition of actual knowledge. Replacing the SAT and ACT will be, essentially, final exams, developed in sync with teacher-selected curricula.

Won’t teachers just teach to the test, which tends to happen with any kind of high-stakes testing? Sure, but that’s what they should do, so long as the tests reflect the ways they normally approach the subject. There may be 10 different calculus tests, reflecting 10 different approaches to teaching that subject. And there could be 50 American history tests, reflecting the different ways teachers might address that very complex topic, focusing on civil rights, for example, or on international relations, or on technological advances.

The tests needn’t be multiple choice, which reflected the technology available in the 1920s, when the SAT was created. Today it’s possible to facilitate the scoring of short answers, essays or even performative responses (e.g. “create a lever to move this crate to that place using these parts”) at reasonable cost; every year will bring more creative ways of testing students’ abilities.

Obviously, the new tests need to be secure. The best way to do that is through transparency: the testing organization should publish every question students might find on their exam. Someone memorizing the answers to thousands of items will score well -- but could have achieved the same results in much less time by simply learning the material. And note that this scale is achievable: artificial intelligence makes it easier every year to create those items.

Finally, good tests don’t make students crazy by appearing to make finer distinctions than they can or should. Reporting scores on a one to 10 or even one to five scale will give colleges all the information they need. And to discourage students from obsessing over the last point, even though that point adds no actual admissions value, we could cut off the top score (leaving us with a one to nine or one to four scale).

So what would a good new testing system look like? Basically, it should be a basket of tests, each created by a textbook company or other curriculum provider in consultation with teachers. The tests would be calibrated to a simple common curve, so a score on one test means the same thing as the same score on another; this way colleges could be agnostic as to which tests students take. Most colleges would require a student to submit three or four scores, and many would require one test to be of English.

Students who register for the tests would have access to the “menu” of possible questions. Testing fees, incidentally, could be paid by the high schools, since the new tests would in some cases replace final exams, as well as the current battery of state tests, most of which share the same flaws as the SAT and ACT.

The new tests will have some resemblance to the College Board’s AP exams. But with different publishers creating different tests, there will be an ecosystem of curricular approaches. The new approach will avoid granting a duopoly, and it will keep prices down and innovation up.

Preparation for these new tests would simply mean learning the curriculum on which they were based. Rather than measure a small part of middle school math and English, these new tests should be connected to a broad range of curricular approaches. And their essays could, perhaps, replace the current, inequitable system of college application essays, which too many people help the student write and edit.

This approach would also address the standardized testing regime’s promotion of inequality. SAT and ACT coaching are best performed by specialized private tutors who understand the tests’ idiom and scope. When tests begin to jibe with school curricula, underwriting test prep for disadvantaged kids would simply mean properly funding K-12 education.

So who will set up this new approach to college admissions testing? California started the conversation, but any state -- or the federal government, a foundation or an association of colleges and universities -- can grab the wheel and steer the college admissions process to fairer and more effective ground.