The Cognitive Science Behind Learning
The most complex thing in the known universe is the human brain. We can change learning practices for the better, and impact outcomes, by understanding cognitive science in general, and the learning sciences in particular.
In most any profession, there is a body of knowledge that drives decisions. Whether it’s medicine-based physiology or flight in physics, practitioners need to understand what’s happening to make appropriate decisions. It’s the same with learning. To be able to determine whether a planned intervention — training, e-learning or otherwise — is appropriate, one must understand learning.
This understanding isn’t simple. The claim has been made, fairly, that the most complex thing in the known universe is the human brain. Therefore, to believe that a systematic and persistent change in operation can be done without a fairly deep understanding of the brain is simplistic. Instead, learning should apply the results of learning science research, a veritable learning engineering. As Will Thalheimer, president of Work-Learning Research and author of “Performance Focused Smile Sheets,” said, “The learning-design process is transformed when scientific research is used.” He cited specific benefits including team direction, innovation and planning.
The consequences of not understanding the brain’s impact on learning, on the other hand, can be costly. Time invested in developing learning that is ineffective is only one aspect. Another is using training or e-learning when other solutions would be more effective. There is a whole suite of solutions currently being sold in the learning industry that have a questionable scientific provenance. “Be skeptical of claims you hear by vendors, bloggers, friends and coworkers. There are innumerable myths and mythologies floating around the learning field,” Thalheimer said. The nuances are subtle. Well-produced experiences aren’t noticeably different from well-designed and well-produced learning, but the outcomes are. Plus, there are some misunderstandings that interfere, for a variety of reasons.
For instance, there’s quite a lot of hype about neuroscience implications for learning, but researchers are quick to point out that most of the important results come from another level. There are levels of analysis; neural is one, but the next level up is the cognitive level. That’s where most of the important implications come from, as well as the social level above the cognitive.
The Cognitive Umbrella
Cognitive science is the most accurate description as it was specifically created as an umbrella term to incorporate all levels of human behavior from neural to social, and it includes contributions from many disciplines including philosophy, anthropology, neuroscience, psychology, sociology and more.
If leaders aren’t aware of the nuances, they can fall prey to some persistent — and expensive — myths. Learning styles, for instance, have been robustly demonstrated to be of no practical value, yet instruments and arguments for sensitive design are still in play. Also, a variety of opportunities to support learning are focused more on aesthetics than effective outcomes. It takes a real understanding to discern the difference between learning and the folk psychology that many people wrongly follow.
“I see a lot of wasted efforts in learning and development that assume that information delivery is going to solve the problem,” said Julie Dirksen, principal of Usable Learning and author of “Design for How People Learn.” Learning leaders need to know more, to do more. Consider the following simplified version of the cognitive science of learning to draw some important implications on what should happen in organizational learning.
Let’s start at the neural level to get down to the basics. At its core, learning is about strengthening the connections between certain neurons. It’s safe to say the neurons that fire together, wire together. Let’s be clear, it’s not like learning strengthens the link between the manager neuron and the leader neuron. The brain doesn’t work like that. Instead, it’s about patterns of activation that represent various things like concepts and actions. Think of a TV set. Different patterns of activation in the pixels that make up the screen create different pictures. Similarly, different patterns of activated neurons represent different ideas and thoughts. When activated together, the links get stronger.
Note, there’s only so much strengthening that can happen at any one time before the brain literally needs sleep. Strengthening takes energy and resources that are depleted and need to be refreshed. Therefore, to make learning persistent, it needs to be spaced, or reactivated and strengthened over a period of time. The amount of time over which to practice, and the total quantity needed, depends on the complexity of the task and the amount of time between practice and performance as well as the time between performance opportunities.
Obviously, if the learning is complex, and there are longer times between performances, more practice is needed. For example, for pilots practicing for emergencies, the amount of practice is ongoing and deep. There we’ve already transcended the neural level and moved into the cognitive level.
At this higher level, learning and instruction is about designed action and guided reflection. Learning leaders should create practice opportunities that require action, and facilitate reflection around that action. Of course, there’s a role for content, providing a framework to guide choices of action in the practice, and to provide a basis for reflection, but the focus is on action.
At the cognitive level, content ideally is a mental model, a suite of causal and conceptual relationships that provide a basis for explanation of what happened and predictions about what will happen. This model, or suite of models, then provides a basis to guide learning actions for reflection, assessing learners’ actions against the model.
One robust finding around models is that learners will build them, and they’re remarkably hard to extinguish if wrong; instead, they get patched. A plausible approach is to make sure there’s a model to begin with, and then refer to that model in examples and in feedback on practice. The alternative is a Frankenstein’s monster of pieces patched together and liable to still have fatal flaws.
Another robust finding is that learners go wrong by bringing in inappropriate models. Most of the mistakes we see are systematic, not random. Therefore, provide the opportunity to make mistakes and to detect and remediate the wrong models in the learning experience before it counts. Also, having silly or obvious alternatives to the right answer doesn’t facilitate learning.
Pattern-matching and Meaning-making
There is some randomness in our actions. However, it’s evolutionarily adaptive. Our cognitive architecture has been successful at pattern-matching and meaning-making, not at the ability to perform by rote repeatedly and accurately. As a consequence, putting rote information in the world instead of the head — using performance support tools such as checklists, decisions trees and lookup tables — makes more sense than courses in situations where the information is dense, arbitrary or changing quickly. Too often learning leaders make courses when the information doesn’t have to be in the head, it just needs to be on hand.
What has to be in the head is the ability to retrieve and apply the information to problems like those learners will face in performance situations. The way to develop that ability is, not surprisingly, practice that requires learners to solve such problems in context. Two critical things here are: tasks that require retrieval of the appropriate knowledge and skills, and creating contexts that resemble the performance situation.
So, create meaningful practice. To train in sales, have salespeople make calls, write proposals and address objections. If employees are in operations, have them trouble-shoot problems, design new processes, write requests for proposals, evaluate responses and so on. What doesn’t work is having them recite back information about these things; they need to actually do them.
Further, part of our learning is aggregating patterns into larger ones, chunking information into coherent wholes. Our working memory, our consciously considered thinking, is limited to only a few chunks. Using a computer metaphor, we compile those chunks, giving us greater facility as we can operate on one chunk instead of several parts. This effectively gives us more operating room.
However, there are several entailments to this. For one, learning should progress in sensible steps and provide sufficient practice to develop the necessary level of performance. These chunks need to be used over time to be automated. If it’s too complex, there’s too much overhead, and the learning isn’t efficient or effective.
Another side effect is that as learning becomes compiled, the information about what we’re doing also is compiled. That is, most people literally don’t have access to what they do. The implication here is that asking subject matter experts to provide the necessary background for learning is fraught with trouble. There are a variety of heuristics for working with SMEs, but requirements that they have to be the source of expertise can be problematic.
At a higher level, one might consider social learning. If a person hears a particular recitation of a concept, they’re likely to take one version of that. When people can experience other interpretations, their learning increases for several reasons. For instance, there’s additional processing that reactivates and strengthens the learning. And, particularly if there’s a requirement to find a resolution among different ideas, negotiation of a shared understanding generates considerable processing and an outcome that’s stronger than any individual component.
Recursively, we can also apply our thinking to our design processes. Just as learning implications arise from our cognitive architecture, so do limits to our design abilities. Externalizing our processes via checklists, for instance, minimizes the problems that occur in random errors. Similarly, templates oriented around the aforementioned principles help keep us from bringing in misconceived ideas about good design. And working together, at least at critical ideation stages, can keep us from a limited experience conceptualization.
The end result is that we can change our learning practices for the better, and impact our outcomes, by understanding cognitive science in general, and the learning sciences in particular. “One of the encouraging things I’ve seen in more effective organizations is programs that are spread out over time so people can practice, get feedback and space out their learning,” Dirksen said.
It’s time to take our profession seriously and avoid learning malpractice. We must not falter, our organizations need us.
Clark Quinn is executive director for Quinnovation, and author of “Revolutionize Learning and Development: Performance and Innovation Strategy for the Information Age.” Comment below, or email editor@CLOmedia.com.