Liftology 9: The role of models

When we talk about learning, frequently we talk about ‘content’. However, that’s too coarse. What we want to do is break that down into the minimum content we need. What is the necessary support for choosing actions? Much of the research points to mental models, causal explanations of how the world works, that let us predict the outcomes of actions. From this basis, we can choose what actions to take. 

The underlying idea, for situations not requiring automated reactions pre-conscious consideration, is that we are choosing between courses of action. What helps us choose is predicting the outcome of action A, and action B, etc. Then, we can compare the outcomes of each, and determine which course of action is optimal. What that requires, of course, is an understanding of how those actions will lead to outcomes.

That’s the role of mental models. They explain how the world works in a causal story of elements and how they’re connected, and what affects those relationships. They’re conceptual, so that they’re not tied to any particular situation, and can be matched. 

Of course, the initial presentation and activities that develop those models matter, but they also need to be reactivated. Research suggests that the best reactivation is to have them used in situations like the ones learners will perform in. In the course of using the models, they have to reactivate them, and that serves the purpose of checking whether they’re there (that is, you don’t need knowledge-test questions to determine that). 

The initial retrieval practice should be accomplished in the learning experience, but as research on spacing suggests, then that retrieval practice needs to happen again and again over time, with challenge increasing,  feedback faded, and interleaving occurring. Of course, we’re also segueing from canned practice to live performance. 

The feedback that one gets should reinforce the role the model plays. The wrong choices typically aren’t random, but made from the wrong or a faulty model. Explicitly, the right choice is reinforced by what the model predicts, and likewise the wrong choices are indicated by the model failure, before the right answer and model being presented. 

The latest cognitive theories posit that we actively build models. Extant research suggests that if we build bad models, we don’t replace them, we patch them. The obvious inference is that the best way to ensure good models is to represent them up front, and then use them to guide learners to become adept practitioners. We can, and should, do so.