What is deep learning?

Our mission at The Learning Code is to help you make your learning meaningful, achievable, and purposeful. We focus our energy on supporting students who want to earn a college degree in the United States. We do so by using three powerful tactics. First, we share with you research-based principles of learning. These principles are founded on cutting-edge cognitive science results to explain how learning works. Second, we feature student stories and battle-tested learning practices that show how to implement these learning principles in your college classes. The learning practices we share with you are designed to inspire you engage in deep learning and also to help you strategically navigate a college education system that is severely flawed. Third, and perhaps most importantly, we work to support you in developing effective help-seeking behaviors so that you can create your own learning teams to facilitate your success in each of your college classes. In this post, we define what it means to engage in deep learning. This post sets a foundation for a later work we will do to explore barriers to deep learning in the US higher education system. All of this will help us develop targeted strategies that you can use to thrive in your college classes. In other words, we are focused on empowering you to become a strategic deep learner and to successfully complete your college degree in spite of institutional barriers that get in the way of your learning.

The phrase deep learning includes exactly two words: deep and learning. To understand what it means to engage in deep learning, let’s break down each of these individually. Then we’ll build a shared vision for what it means to be a deep learner. A great place to start this journey is to define what learning is.

What is learning?

DEFINITION: LEARNING (noun)

Let’s define learning as a growth process that happens inside your body and leads to changes in your knowledge, beliefs, behaviors, or attitudes. These transformations occur based on your experiences and increase your potential for improved performance and future learning (adapted from How Learning Works: Seven Research-Based Principles for Smart Teaching by Ambrose et al.).

Some features of this definition are worth exploring:

  1. Learning is a growth process that takes place in your mind and in your body. Learning literally involves activating brain cells, growing new connections between brain cells, or strengthening existing neural pathways so that the neural networks that you’ve already formed become more robust, durable, and lightening-fast.

  2. Learning involves change in your knowledge, beliefs, behaviors, or attitudes. When you engage in the process of learning, you create new capacities within your body and mind. This creative process includes two fundamental activities. First, you can adapt new functional properties of neurons and develop new neural pathways. Second, you can build stronger connections within an existing neural network.

  3. As you learn, your body and your brain physically change. These changes are designed to capture, hardwire, and biologically encode your learning inside a vast network of cells that run throughout your brain and your entire body.

  4. Learning is a process, not an isolated event. In other words, the changes that occur as you learn unfold over time and cannot occur in an instant. To learn and change, you must repeatedly engage in practice over many days, weeks, months, years, or decades.

  5. Learning happens based on your experiences and results directly from what you do and what you think. The learning process is something that you do for yourself. Learning is not something that others can do for you or to you. The only way a teacher can advance your learning is by influencing what you do to learn. This is very similar to building athletic skills. An experienced coach can help you decide on which workouts and training activities are most appropriate for your development. However, to grow your athletic skills, you are the one who has to workout, train, and practice. The same is true for growing your intellectual skills.

  6. You can improve your capacity to learn and you can grow your learning abilities. To do so, focus on building your learning strategies. Seek out accurate information about how learning works. Look for efficient and effective learning practices to acquire new information and construct new knowledge. Also be on the hunt for practices that leverage active and attentive repetition to strengthen your ability to recall ideas that matter to you. Finally design reflective learning habits that empower you to identify your strengths and weaknesses, correct mistakes, and think flexibly about the way you perform tasks you are trying to master.

This definition of learning highlights some of the most important features of learning. However, not all learning is created equal. Our goal at The Learning Code is to help you develop habits of strategic deep learning. Now that we know a little about what learning is, let’s explore what it means to learn deeply.

What is deep learning?

DEFINITION: DEEP LEARNING (noun)

Let’s define deep learning to be learning that involves an intense, distraction-free focus on growing your abilities by pushing beyond the limits of your current capacity. When you engage in deep learning, you actively reach for and repeat skills that you want to build by paying extra special attention to your performance during each repetition. (adapted from The Talent Code: Greatness isn’t Born, it’s Grown and The Little Book of Talent: 52 Tips for Improving Skills, both by author Daniel Coyle).

Notice that in this definition, we focus on two distinct features of deep learning:

  1. Deep learning requires that you seek out a feeling of heightened struggle as you reach towards a goal that is on the outer edge of your current abilities. To learn deeply, you must exert conscious effort to push yourself into a space where you make frequent mistakes. Deep learning also requires that you actively reflect on each attempt, correct your mistakes, and work to improve your performance. Thus, when you learn deeply, you want to be careful not to push yourself too far beyond your current level of mastery. You want to maintain your ability to identify, recognize, and reflect on your mistakes as they happen.

    When we look back to our definition of learning as a growth process that happens inside your body, we can say that the first feature of deep learning involves creating neural networks in your brain to encode the knowledge or skills you want to build. Your brain is designed to capture this effort by making new interconnected pathways between neurons or by altering and adapting the function for neurons not previously used for this purpose.

  2. Deep learning also focuses on active and observant repetition. Not only are your trying to develop new knowledge or learn a new skill, but you’re also working to build mastery so that you can leverage this knowledge with progressively less effort. The first stage of deep learning involves successfully mastering your target ability for the first time. This second part involves repeating the process of reaching and performing so that you can leverage your new ability more quickly. Each time you repeat, you give yourself another chance to make mistakes, identify and correct your errors, and to refine or expand your capacity with respect to your new ability. This vigilant, repetitive practice solidifies your learning and gives you faster access to your new skills in your future.

    The second feature of deep learning is designed to strengthen a nascent neural network that exists in your brain. This process involves reinforcing recently-created neural pathways by building myelin sheaths throughout the network. These myelin sheaths speed up signal propagation and thus allow you to activate your new ability more quickly with less effort.

Now that we have introduced a definition for deep learning, let’s expand on our understanding of what it means to learn deeply. Specifically, let’s define the opposite of deep learning, which we will call shallow learning.

What is shallow learning?

DEFINITION: SHALLOW LEARNING (noun)

Shallow learning is the opposite of deep learning. Shallow learning involves a modest level of focus or lacks the intensity and duration required for deep learning. When you learn in a shallow way, you probably do not have a clear vision for how or why this learning is important to you. You may also be unwilling or unable to leave your current comfort zone or to reach for goals on the outer edge of your ability. Shallow learning involves a distaste for mistakes and a desire to avoid making or identifying errors.

The special features of this definition include:

  1. Shallow learning is marked by the absence of a conscious effort to engross yourself in intense focus on hard tasks. When you perform shallow learning, you may be agnostic about or disinterested in protecting yourself from distraction. You may even be exposed to environments that require lots of multitasking as you try to learn. For example, think about trying to read while watching TV or trying to listen to a lecture while texting with a friend. This lack of deep focus makes it very hard to engage in skill development beyond the frontiers of your current levels of mastery. While you may be able to complete the task you are working on, you invest limited conscious effort on your learning. Thus, the skills you build using shallow learning are encoded in a sparse neural network and are quickly forgotten once you end your learning session.

  2. Shallow learning involves completing tasks that you do not feel strongly about. When learning in a shallow way, you probably have weak, unconscious, or even negative feelings about the learning activity you are working on. You likely do not have a strong internal desire to learn and you are probably not excited about or deeply engaged in your learning. You might not see this learning as linked to your identity nor can you articulate how this learning contributes to your long-term vision for your future. You might even be motivated by fear rather than desire if you perceive that you are required to do this learning by an authority figure who does not know you and with whom you do not share an authentic, trusting relationship. In any case, you are not excited by the learning, you are not struggling, and do not feel that putting in the extra effort to challenge yourself is worth your energy.

  3. Shallow learning involves using strategies that do not support deep learning and thus slow the speed at which you develop new skills or create new knowledge. As we discussed in point 6 under our definition of learning above, not all learning practices inspire learning equally. When engaged in shallow learning, it may be the case that the practices you are using to learn do not sufficiently challenge you to learn deeply.

For the rest of this post, we’ll explore some learning strategies you can use to engage in the process of repetition to inspire deep learning and avoid shallow learning. The question of how to build knowledge effectively for the first time is a much more challenging question that we’ll address in the future.

What strategies support deep learning?

Suppose that you want to to strengthen an existing neural network using repetition to inspire deep learning. A natural question that arises is: what type of learning strategies are effective for repetition? A related question is: what type of repetition strategies result in more shallow learning? Below, we explore some effective research-based strategies for using repetition to produce deep, durable learning. We also highlight some strategies that result in shallow learning. All of this work is based on the book Make It Stick: The Science of Successful Learning by Peter C. Brown et. al.

Focus on effective strategies for deep learning

  1. Retrieval practice: recalling facts, concepts, or techniques from memory.

    The act of retrieving knowledge from memory has the effect of making that knowledge easier to recall again in the future. To be most effective, retrieval must be repeated again and again, in spaced out sessions, so that the act of recalling knowledge requires cognitive effort rather than becoming a drill in mindless recitation. The more cognitive effort required for retrieval, the greater your retention will be.  

  2. Distributed practice: space out practice sessions over many days, weeks, or months

    Practice is far more effective when it’s broken into separate periods of training that are spaced out. When learning something new, create learning habits in which you practice over many days, weeks, or even months. Ideally, you might create training periods that occur on a daily basis spanning many weeks.

  3. Interleaved practice: mix different ideas together as you do you retrieval practice, each time in a different order

    When studying a set of ideas, learning is much more effective if you interleave different ideas during your study. To do so, you want to work on multiple skills at the same time. Contrast this with block practicing during which you focus on only one skill at a time.

    When you engage in interleaved practice, be careful not to focus on component skills in discrete chunks. For example, don’t practice one definition for 30 minutes, then moving on one theorem for the next 30 minutes, and then finish by analyzing one example for the last 30 minutes. Instead, mix each component skill together during your practice. First test yourself on a definition, then solve the problem, and then state a theorem. Repeat this cycle in a different order. Mix in materials from previous classes and reshuffle each idea so that you never study the same material in the same order. Interleave your practice so that each time you retrieve an idea, you have to activate your memory and think critically about what you are doing.

  4. Varied practice: frequently change the tasks you are working on.

    During your learning sessions, change your tasks frequently so that you constantly confront novel instantiations of the knowledge you’re trying to build. Instead of solving a single problem or answering the same question multiple time, include many different problems that focus on the topic you want to learn. Try to choose diverse problems that test different aspects or applications of your target idea.

    A great example of this is described in a famous bean bag study in which a group of eight-year-old children practiced tossing beanbags into a bucket in gym class. Half of the kids practiced tossing bags into a bucket three feet away, never varying the distance of the bucket during their practice (not varied). The other half of the children practiced tossing their bags into buckets that were either two feet away or four feet away, where the distances changed during each practice session (varied practice). After twelve weeks, the kids were tested on tossing a bag into a bucket that was three-feet away.

    The top performers were those kids who engaged in varied practice on buckets that were either two or four feet away. This was true even though these students never practiced on a bucket that was three feet away. By having to vary the type of practice they did, they developed a more diverse set of muscle controls and proprioceptive skills.

    When you vary your practice, you develop skills and knowledge targeting different dimensions of the knowledge you are trying to build. This leads to a richer and stronger neural network and thus inspires deeper learning.

  5. Embrace difficulty: when learning is harder, it’s stronger and lasts longer.

    There is an inverse relationship between the ease of retrieval practice and the power of that practice to entrench your learning. The easier it is for you to retrieve your new ability, the less your retrieval practice will benefit your retention. Conversely, the more effort you have to expend to retrieve your desired knowledge or perform your new skill, the more the retrieval practice of will result in long-lasting learning.

Avoid strategies that result in shallow learning

  1. Avoid block practice

    Block practice is the rapid-fire repetition of a single skill or idea you are trying to burn into your memory. This is the conventional wisdom of “practice-practice-practice” on a single idea or skill until you’ve fully mastered that skill.  

    A great example of this would be to spend an hour focused on trying to learn a single definition from one of your courses. During that time you focus only on that idea. While blocked practice seems enticing as a strategy, it is actually a very inefficient method of committing an idea to memory. This is partly because blocked practice doesn’t sufficiently challenge your brain.  

    If you test yourself on an idea and then retest yourself two minutes later, you haven’t given yourself a chance to forget what you just studied. When you re-study that idea after only a small amount of time, you are not creating new neural new pathways or strengthening your long-term memory. Instead, you are mindlessly regurgitating the idea you just studied. Remember, the harder something is to do, the more likely it is that you are engaged in deep learning.

  2. Avoid rereading text

    Rereading text is the process of repeatedly reading your lecture notes, textbook, or homework solutions. This also includes the process of re-watching an online video or highlighting texts that you’ve already read. This method of study is unproductive from the standpoint of deep learning.

    When you reread a text, you are not actively testing yourself on the ideas from that resource. Instead, you are heavily relying on the author’s knowledge to guide your thinking without actively producing thoughts of your own. This is analogous to working out at the gym using a spotter. When you re-read text, you are effectively lightening the load by having the author carry most of the heavy weight. Don’t spend your precious time on this relatively mindless activity. Remember, deep learning is hard and should require active engagement.

  3. Avoid massed practice (i.e. avoid cramming)

    Massed practice is learning that occurs with no intervals or short intervals between successive bouts of learning. Such practice is designed to complete many hours of studying in one session. This might be followed by a short break and many more hours of studying on the same topic.

    A great example of massed practice is the strategy of cramming in which you wait to study until the days right before a major due date. Then you binge study and try to learn all the content at once. This strategy sometimes produces decent tests scores but does not result in durable learning. In fact, while you may be able to recall your learning in the few days after a massed practice session, these ideas are quickly forgotten and you need to relearn this knowledge if you want to use it in your future.

    This strategy is very inefficient. You aren’t spending enough time to encode the ideas in your brain, make mistakes, correct your mistakes, and then try again over many days. Massed practice results in shallow learning that makes it harder to remember the new knowledge you targeted far into your future.

  4. Avoid one-off problems

    One-off problems are problems that you solve once and never think about again. Many times when solving a one-off problem, you might write your solutions on scratch paper, fail to methodically track your work, or never create a system to capture the thought process that lead to your solution so that you can access your work at a later date. Instead, you care only about the answer to the problem. Once you get your answer, you move on and don’t look back.

    Many young math or engineering students view assigned homework problems from this paradigm. The thinking goes like this:

    I want to earn my college degree. To do so, I must complete this class. I want to succeed so I will try to get an A in this class. My professor says that my grade is based on graded homework and three exams. In order to get an A in this class, I will try to get an A on all my homework assignments and on my exams. To get an A on my homework, I have to get the right answer to all to the problems on my assignment. Because getting an A is what is important, I need to focus on getting the right answers to each of the problems. As long as I get the right answers, I am going to succeed on my homework. With this type of effort, I will most likely also succeed on my exams and in this course. If I just repeat this strategy every quarter, I will earn my degree and go on to have a very successful college career.

    From this perspective, it is easy to understand the feeling that a problem can be discarded after you’ve found the right answer. Following this logic, it makes sense to solve a problem once and never engage in retrieval practice or repetition. However, this is a less-than-useful strategy. Remember, just because you understand doesn’t mean you can recall or remember, especially under pressure. One-off problem solving is a classic example of a focus on transient understanding rather than durable learning. Sadly, many college teacher create policies that incentivize this type of short-term thinking instead of a focus on deep learning.

We’ve now defined learning, deep learning, and shallow learning. We’ve also explored some useful strategies to engage in repetition as part of deep learning. Throughout this discussion, we’ve deferred a meaningful analysis of how your level of motivation relates to the energy you dedicate towards deep learning. We’ll also put off analyzing how systems of oppression and socioeconomic factors effect a student’s ability to learn deeply within the US higher education system. We will address these topics in later posts on The Learning Code. Let’s end with some useful reflection questions on the terminology we’ve developed in this work.

Community Challenge:

  1. Come up with a first draft of your own definition of learning. Capture your ideas and write your definition in a special place so that you return to and edit your ideas in the future. As you develop your work, engage in retrieval practice by trying to recreate some of what you read in this blog post from memory.

  2. Write a first draft of your definition for deep learning.

  3. Draft a definition for shallow learning.

  4. Describe what it feels like when you engage in deep learning. What type of subjects and topics do you already do learn deeply about?

  5. Describe what it feels like when you learn in a shallow way. When do you tend to engage in shallow learning? What factors in your life and what type of classroom policies tend to make you focus on shallow learning rather than deep learning?

  6. What type of challenges do you face when you try to transition away from shallow learning into deep learning for a topic that you are just starting to learn about?

  7. Use your own words to describe each of the five most effective strategies suggested in this article. Try to recreate what you read from memory. Seriously. Put a timer on for 5 minutes and try to write as much as you can about the strategies that empower deep learning. When the time goes off, go back, read through that section of the article again and correct your mistakes. Then cover your work, and try to do that again.

  8. In your own words, describe the four less-than-effective strategies suggested in this article. Leverage retrieval practice to test your memory and try to build your ability to remember each of these strategies.

15 thoughts on “What is deep learning?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s