Testing Responsibility

You may remember that reader Loki on the run wrote:

We may have spent a hundred years investigating how people learn, but the best way to learn to ride a bike is to get on one and try, and to pick yourself up when you fall off and try again. Having a parent run along behind to hold the bike up is good at first, as are trainer wheels, but eventually, you have to spend time riding the damn bike.

It seems to us that many (perhaps most) students today have been given the idea that they have no responsibility to learn and that teachers have all the responsibility for their failure to master the material. They believe in instant tratification [sic] and will not put in the time with the homework and the exercises. That is, they will not ride the bike and expect to become BMX celebrities simply by being told about angular momentum and bearings and friction.

Loki is, to some extent, right. If students don’t take responsibility for their learning, then there’s no hope. Despite the teacher’s best efforts, a kid who’s bent on shirking the material won’t learn it. The old adage “You can lead a horse to water” comes to mind. But perhaps Loki is being a little too hard on the students, on the teachers, on everyone. Still, it’s difficult to know what Loki means by responsibility. You might be suprised that incentives (such as money or the promise of a class party) are less effective at bolstering performance than really explicit directions and prompts. (Don’t believe me? I’ve got references.) So maybe we should at least hold teachers responsible for letting the students know what they’re responsible for.

That said, I’d like to acknowledge that people can do more with the help of others than they otherwise could alone. Some psychologists have studied this phenomenon formally. They’ve identified a zone of proximal development (ZPD). The ZPD is something like the teaser accompanying the end credits of a television show that gives some hint as to what will happen next time. The very existence of the ZPD shows that learning is necessarily social. Or at least, effective learning is social. I’m not going to argue that people cannot learn alone. But we’re talking about building effective classrooms. Let’s not make it harder for the students just because we can. So, to use Loki’s metaphor, it’s very useful to have training wheels and parent nearby. On this neither of us disagrees. The problem comes in when we try to decide what the parent (or teacher) ought to do.

In the model which Loki presents as standard, it seems first the teacher presents a repository of knowledge—very likely in the way of facts and procedures, e.g., the product of two negative real numbers is a positive real number, or the algorithm for multi-digit addition—then the students memorize and reproduce the facts and procedures. Teachers evaluate the degree to which students have mastered the material by way of tests. It is very likely that these tests ask the students to answer questions written in a format consistent to the teacher’s original presentation. To perform well, the students need to memorize and drill until their responses become automatic. This form of evaluation suffers from at least one critical problem: it cannot distinguish between accurate performance and thorough understanding.

The performance of a good novice and an expert can often appear the same. For example, a child who simply learns his addition tables by rote can respond as quickly and accurately as another child who has a reasonable grasp of the mechanism represented by addition. Thus when the two students move onto problems which require a “carry,” the first student will have a significantly harder time simply because he has more facts to memorize, whereas the second student will be able to generalize the rules of addition to accommodate the new problem.

I’ve discussed test design before, but for Loki’s benefit maybe I should quickly recapitulate. A Good Question should be able to distinguish between accidental correct answers due to rote memorization and intended correct answers resulting from mastery over the subject. Let’s build up a good problem from a bad one. When learning about prime factorization, teachers often introduce the concept of the least common multiple (LCM) and greatest common factor (GCF) of two numbers. Therefore, a natural question to ask on a test might be:

Standard Question. Find the least common multiple of 12 and 21.

In itself, there’s nothing especially bad with the Standard Question. It gets to the point, shows that the student has some computational understanding of what’s going on, and can reliably produce the answer to this type of question. In fact, a Good Question draws on the content of interest. If we’re interested in LCM, then this question is on its way to becoming a Good Question. But if the student taking the test has access to a TI-89 or other sophisticated calculator (as I did), then all he needs to do is to type LCM(12, 21) into the calculator. Surely, the use of technology is not something to be scoffed at. I’m using a computer to type up this paper, after all. I’m not about to propose everyone throw out their computer and write everything by hand. But if our aim is to teach kids something about the structure of numbers, then maybe a heavy dependence on technology gets in our way. We really need to come up with a Better Question, one that a calculator can’t do. Let’s try.

Better Question. Tricia says that you can find the least common multiple of two numbers by finding their product and dividing by their greatest common factor. Does Tricia’s method always work? Explain your answer.

Well, we’re getting there. Except now Loki might object, and rightfully, that this Better Question doesn’t readily test whether students can “ride the bike.” It asks them to identify the various parts. It even requires them to be able to build the bike. But it doesn’t ask them to ride it. So, maybe a Good Question does it all: it requires kids to build and ride their own bike. What more responsibility could we ask for then that?

Good Question. Tricia says that you can find the least common multiple of two numbers by finding their product and dividing by their greatest common factor. Does Tricia’s method always work? Explain your answer. Find the LCM of 12 and 21 in at least two different ways.

And notice that the Good Question requires students to calculate the LCM in at least two different ways—here we sort the lazy memorizers from the more dedicated kind. What makes the Good Question good, though, is that it asks the students to synthesize knowledge on the spot. That’s not a skill you can easily happen on by mistake. Sure, it’s a little bit harder to grade, but who cares; isn’t that the point of being a teacher?

As a test writer, I see myself in a very funny and useful position. Teachers have a habit of “teaching to the test.” So if I alter the way I write tests, it seems—at least in theory—that I accomplish real change in the way teachers prepare their students. Ideally, teachers would have enough mastery over their subject so that they could let students lead the learning themselves (as is done in the Math Circle run by the Kaplans at Northeastern and Harvard, or in schools which have adopted a curriculum tailored by Project SEED). In those classrooms, the shift in responsibility is more apparent—though perhaps no more real, since the teacher must keep a careful eye on the course of the class and give constant, mindful guidance. Perhaps this is more what Loki had in mind. I’m not sure; hopefully, he’ll elaborate. For now, I feel like I’m working on both the teacher and the student in a way that produces a broad effect on practice without having to sort through the politics of education policy.

In my next response, I’ll address the social component of learning more directly. Sorry guys, this post went in a different direction than I had initially intended. If I don’t use the words authoritarian and authoritative in my next post, please leave me an angry comment.


See, for example, Carroll, W. R., Rosenthal, T. L., & Brysh, C. G. Social transmission of grammatical parameters. Psychological Reports, 1971, 29, 1047–1050.

Rosenthal, T. L. & Zimmerman, B. J. Language and Verbal Behavoir: Social Learning of Synactic Constructions in Social Learning and Cognition, Academic Press: New York, 1978.

Technorati Tags: , , , , , , , , ,