Power of Prediction

We predict things all the time (usually based on prior knowledge, context, and experience). Have you ever attempted to complete someone’s sentence? But did you know predicting in a novel context can be a potent learning strategy? Consider the following code segment:

var counter = 0;
while (counter < 10) {
  print(counter);
  counter++;
}

Even if you have no experience in programming, look over that code and make a prediction about what might happen. Think about what var might mean. What do you think “while counter < 10” might do? What does print do? What do you think the final value of counter might be?

You may have just skimmed over the code segment above. But maybe you stopped for a twenty seconds and pondered what some of those lines of code do – and if you did invest a few seconds puzzling over the code, here’s what’s going on in the program:

1. We establish a variable​ called “counter” to have a value of zero.
   
2. Then we look at the value of counter. If it’s less than ten, we will output the value of counter via a print statement (so the first time the value is output, it ​will be “0”). Next, we’ll increase the value of counter by one (that’s what the ++ does). And again, we’ll examine the value of counter. If it’s less than ten – which it is because counter is now one – we will output it again and increase it by one again. In fact, we’ll do that a few more times until the value of counter is no longer less than ten.
   
3. The program has now ended because counter is no longer less than ten.

It’s likely you were able to predict at least some part of the code (for instance guessing that “var” meant “variable” or figuring out that counter started at 0 and ended around 10). Now go back and examine the code segment again. What parts did you get right? What parts now make sense after the quick debrief?

The goal of this exercise isn’t to teach you programming though; the intent is to demonstrate the power of prediction. 

I like using prediction exercises with my students because I find it promotes inquiry (one of FLCC’s values to boot!) and curiosity. I like giving students the opportunity to wonder. But there is a much more potent explanation for why prediction works; in When generating a prediction boosts learning: The element of surprise (Brod, Hasselhorn, and Bunge, 2018) the authors conclude that incorrect prediction leads to surprise when learning the correct answer (which strengthens learning):

In conclusion, this research demonstrates that there is a specific benefit of generating a prediction for learning, and that this benefit is at least in part mediated by the surprise generated by expectancy-violating events. It presents convergent evidence from several approaches – in this case, behavioral and psychophysiological – and has theoretical implications for our understanding of human cognition as well as practical implications for the development of good instructional practices and study habits.

In Small Teaching by James Lang, an entire chapter is dedicated to prediction. Lang starts with the assertion that “making predictions about material that you wish to learn increases your ability to understand that material and retrieve it later,” (p. 42) and devotes the rest of the chapter providing studies and examples that support this. Even pretesting serves a valuable cognitive function (p. 44):

“…giving students a pretest on topics to be covered in a lecture [asking learners to give answers to questions or anticipate outcomes about which they do not yet have sufficient information or understanding] improves their ability to answer related questions about those topics on a later final exam.”

In Five Teaching and Learning Myths-Debunked by Brown and Kaminske, the authors suggest that prediction also leverages retrieval (another very potent learning mechanism) for students as they continue their academic journey in a specific discipline. Recall that retrieval practice is the (oftentimes uncomfortable) action of recalling previous information – which actually strengthens the memories as a consequence – and adds to the coherence of the memory.

While most people might think that memory is used for remembering the past, learning and memory researchers have actually argued that memory helps us predict the future. Our memory system takes into account complex patterns based on the environment, context, and frequency, that help generate the best possible target with the information it has.

This is also noted in Superlearning by Peter Hollins (p. 21):

A bigger narrative helps contextualize these facts and makes them mean something. In this context, the brain is more likely to retain the information it actually needs to know about the subject. You would be able to essentially predict the facts with a reasonable degree of accuracy once you understand the underlying concepts and how they interact. True, you may not have “memorized” certain information, but when necessary, you can logically work your way through the question and arrive at the same answer as if you had memorized it. This is known as concept learning. It shows us how to categorize and discriminate items according to certain critical attributes. It entails pattern recall and integration of new examples and ideas. And rather than being a mechanical technique of grinding memorization, concept learning is something that must be constructed and cultivated.

Of course, there is also collateral learning as a consequence of leveraging prediction: being wrong. Normalizing predictions (regardless of if they are accurate or not) attenuates the feeling of missing the mark. And that’s great – people need to fail, fail often, and fail gracefully to learn. I suppose constant prediction also helps students refine their skills in formulating hypotheses, too. Making predictions and then discussing them is a brilliant manifestation of metacognition – what we know to be one of the best tools for learning.

One last thought about prediction – a search at Google Scholar will surface an overwhelming number of research regarding prediction as a mechanism for artificial intelligence learning. Prediction is a tool that is leveraged frequently in the domain of AI. So much so that there is research being conducted (Choosing prediction over explanation in psychology: Lessons from machine learning by Yarkoni and Westfall) indicating that prediction might have a place in modern psychology:

We propose that principles and techniques from the field of machine learning can help psychology become a more predictive science. We review some of the fundamental concepts and tools of machine learning and point out examples where these concepts have been used to conduct interesting and important psychological research that focuses on predictive research questions. We suggest that an increased focus on prediction, rather than explanation, can ultimately lead us to greater understanding of behavior.

If you’d like to learn more about prediction, I predict you’ll find Small Teaching by James Lang a great starting point.