This is the third in a series of posts on the book Small Teaching: Everyday Lessons from the Science of Learning by James M. Lang.

Chapter 2 - Predicting

The second Small Teaching practice shows how prediction improves retention and comprehension. It starts by introducing a few research studies, including one very elegant (and surprising!) one from Kornell, Hayes, and Bjork dx.doi.org/10.1037/a0015729. They asked students to memorize pairs of loosely connected words (‘olive–branch’, ‘mouse–hole’,‘whale–mammal’). One group got 13 seconds to study the pair, while the other had 8 seconds with the first word only, then had to predict the second word, then got 5 seconds with the word pair. When tested on how well they could recollect the second word, given the first word, the second group significantly outperformed the first group. Note that this isn’t the kind of test where careful thought would lead to better predictions (‘whale’ could easily pair with a huge number of possible words). Students learned better when they had to predict the answer - even if it was the wrong answer.

One reason for this touches on an important difference between knowledge in experts vs novices. Experts have a dense network of connections between facts, while the knowledge of novices is sparse. When adding new facts to this dense network, an expert makes multiple connections and more easily sees how to use the new knowledge. Learning using prediction works much the same way, activating prior knowledge, and adding the new information tightly into the network of information we already have. It may also help us to see the gaps in our knowledge, through wrong prediction (‘Illusions of fluency’ is the term for what this exposes. I think we’ve all been there.)

In the context of a Carpentry lesson, we can think about using prediction when introducing a more complex construct (‘what happens when we combine grep and wc together?’), when we are modifying an existing problem (‘how will the graph change if we make this change to our matplotlib code?’), or when introducing a new concept (‘how can you imagine using version control in your current work?’).

No matter the way you use prediction, you want to make sure that the learners get the right answer fairly soon (not much of a concern in a Carpentry-length lesson), that you provide time for students to ponder (it is tempting, when rushing through a lesson, to not wait long enough), and that you build on existing knowledge (not asking for predictions before you have provided enough context).