Minimum Standards Testing for the AI Era

The AI era has drastically increased the number of students who are learning nothing at all. A new kind of testing, which I call Minimum Standards Testing, can help discover those students, reduce AI cheating, and allow schools to respond appropriately.

Today’s tests are focused on measuring levels of subject mastery and problem-solving ability in ways that help differentiate A, B, and C level students. But there are no tests that differentiate students who learned the bare minimum from students who learned nothing at all. In other words, there are no tests designed to detect which students should fail and repeat the course.

Such “Minimum Standards” tests would look very different from usual tests. Questions wouldn’t have varying levels of difficulty. They would all have the same level of difficulty: very, very, very easy. A students, B, students, and C students would all get perfect scores on Minimum Standards Tests. These tests would only help to discover which students had learned nothing, not even the barest basics.

Failing a Minimum Standards test should lead to immediate intervention. That could involve intensive tutoring and/or repeating the course. 

Creating Minimum Standards tests will require teachers to design tests very differently from what they’re used to doing. Rarely do teachers ask, “What is the easiest possible question for this course?” How often do American history teachers ask who the first president of the United States was? How often do calculus teachers ask students to find the derivative of x² on a final exam?

Minimum Standards tests can be announced or given as surprise evaluations. Importantly, grades on Minimum Standards tests should not be averaged into final course grades. Making Minimum Standards test scores part of a student’s grade would lead to even more grade inflation than what we have now. Instead, they should only be used to detect dire situations that require immediate and drastic intervention. Any normal student should get a 100% on a Minimum Standards test. Students who don’t should receive immediate intervention.

Outside of academics, minimum standards tests are common. Hearing and vision testing doesn’t help find who has the best hearing and eyesight. It is only there to find major problems that need to be addressed. Minimum Standards testing within courses should operate on the same principle.

Minimum Standards tests should consist of somewhere between one and ten problems. The problems should be so easy that if a student gets a single one wrong, the teacher is willing to fail the student and having him repeat the course. Instructions should be simple. There should be no tricky questions. The test should be such that only a student who has utterly failed to learn the most rudimentary parts of the course could possibly get it wrong. 

AI cheating is creating untold numbers of students who are learning absolutely nothing. Minimum Standards testing can help discover those students and allow schools to intervene appropriately.