Peer Instruction: Engaging students one-on-one, all at once

This paper by Catherine Crouch, Jessica Watkins, Adam Fagen, and Eric Mazur looks pretty exciting to me:

Peer Instruction is an instructional strategy for engaging students during class through a structured questioning process that involves every student. Here we describe Peer Instruction (hereafter PI) and report data from more than ten years of teaching with PI in the calculus- and algebra-based introductory physics courses for non-majors at Harvard University, where this method was developed. Our results indicate increased student mastery of both conceptual reasoning and quantitative problem solving upon implementing PI. Gains in student understanding are greatest when the PI questioning strategy is accompanied by other strategies that increase student engagement, so that every element of the course serves to involve students actively. We also provide data on gains in student understand-ing and information about implementation obtained from a survey of almost four hundred instructors using PI at other institutions. We find that most of these instructors have had success using PI, and that their students understand basic mechanics concepts at the level characteristic of
courses taught with interactive engagement methods. Finally, we provide a sample set of materials for teaching a class with PI, and provide information on the extensive resources available for teaching with PI.

Their stuff is all about physics. I’d like to do it with statistics. I think it could revolutionize the (currently crappy) state of statistics instruction.

4 thoughts on “Peer Instruction: Engaging students one-on-one, all at once

  1. Why is the current state of statistics instruction crappy? Would be very curious to know what you think (perhaps you blogged about this before?) I am new to Bayesian inference and statistics so this is very topical for me. So far I have been your book, BDA, very helpful.

  2. Years ago, I took an evolutionary genetics class that used this Socratic Method. It was fabulous, and I leared a great deal more than I would have otherwise. I think there's a reason they still use this in law school, and perhaps it should have more of a place in higher education.

  3. Andrew,

    I would really like to hear more about what you have in mind. Have you written about this elsewhere?

    It's also interesting to hear the Socratic Method mentioned outside of legal education. I work as an educational technologist at a law school and discussions about the Socratic Method and its strengths and limitations is a constant theme among my colleagues. What I most often hear is that the Socratic Method must yield some of its ground in law schools to other instructional methods.

  4. Thanks for pointing me at an interesting paper. This looks similar to some of the ideas presented in Black and Wiliam (1998) "Inside the Black Box" and stuff that Jim Minstrel and Earl "Buzz" Hunt have been working on for quite some time. Black and Wiliam talk about "hinge questions" which seem to be similar to the Concept Tests in the paper. Minstrel and Hunt advocate "benchmark tests" that occur before the topic is introduced, to help students start thinking about the problem.

    I was reading some of Dylan Wiliam's papers while preparing for a recent tutorial on Bayes nets. I got inspired and added a few benchmark and hinge questions. Stretching what was at one point 1.5 hours of lecture into two hours including the problem sets. I was reasonably pleased with the results.

    1) Benchmark Problem — Before introducing the idea of conditional independence, I had the students work the "Accident Proneness (Spurious Contagion)" example from Feller (1968). I recommended the students discuss their answers with their table-mates and wandered around the room for a few minutes, peering over shoulders and answering questions. I then pulled it up a level to a general class discussion (there were about 20 people in the room) and introduced the idea of conditional independence.

    I think this worked really nicely, because the problem is very difficult to solve without a good understanding of conditional independence. After introducing conditional independence (and a graph to explain it) it no longer seems so mysterious. Thus, why they should pay attention was really underscored.

    2. Hinge Question. Here I was working on the topic of d-separation (the rules from Pearl, 1988, about how to mark conditional independence in a graph). This can be pretty occult, but the students were nodding at the lecture as if they understood it. "Great," I said, "you look like you understand it. Try this problem, and discuss it with your table-mates. I'm getting a glass of water." After a few minutes of letting them struggle with it, I solved the problem at the front of the class.

    This was really great, as I think it really solidified the understanding of the students (many of whom were experienced professionals and associate professors) and helped them understand what they did and did not know. I think they asked a whole bunch of clarifying questions they didn't have before they tried the exercise.

    I'm willing to try it again. Actually, the next time I'm invited to give a one hour lecture, I'm thinking of putting a hinge problem in at about the 30 minute mark. Wake the audience up and force them to actually use the material instead of enjoying my research like a baseball game.

Comments are closed.