Deep Learning vs. Surface Learning

It occurred to me while I was practicing various coin tricks that there is no possible way that these movements could be learned by studying theory alone. For some skills, surface learning is not always possible. As an example, to pass a coin magic test, a bare bones demonstration of the manipulations based purely on memory would not suffice. Even if your test were just to demonstrate deliberate movements to a master magician, but not at a level of proficiency to fool even an inexperienced observer of magic, but rather to show that you had been practicing (doing the thing), and all you had done was cram the theory in the night before, you would fail miserably. The slow, deliberate, mechanical movements that might be required for such a test would not even be possible by knowing just the theory by memorization alone. Some skills require doing, plain and simple.

Then I started to think about other learning experiences that do not involve physical manipulation, yet still require the same level of dedication to doing in order to pass a test. Something that I learned very early on in my academic career was that some skills require doing in order to even pass the class, let alone do well in any sense. For example, in my Math 101 class (Calculus 2), the failure rate was 50%. Half of the students taking the class were taking it for a second time. Yes, there are gifted mathematically minded students who think that math is a breeze, and they never really had to try very hard. However, for the average person, the only possible way to pass classes like that are to do problems ad nauseam; to spend literally four hours per day, nearly every day for the whole semester was the only way to get a good mark at this class (unless you were one of the gifted few). So, learning by doing isn’t a requirement to learn only physical manipulation skills. Some subjects require a deep understanding in order to even pass the class.

Many teachers of these subjects simply present the material, expecting that their students will pass the course or they won’t, and that’s the students responsibility. In a course similar to Math 101, Linear Algebra was another tough course (at least for me). I remember going to office hours where my professor refused to show me an example of how to go through a sample question, stating that I would just memorize the steps in order to complete similar questions in the future. I got the distinct impression that this professor did not care about how his students did. According to John Biggs, who created the SOLO (Structure of the Observed Learning Outcome) Taxonomy, the highest “level” of teacher are the ones who care about how students do during and after the lecture material, and establish outcomes of learning for the course, and encourage engagement with the material. Clearly, this was an example of poor teaching practice.

If I was to be a a level 3 teacher, one who cared about how students did during and prior to a lecture, and I was my own student, how could I teach myself to be a better coin magician? First off, I would establish a set of learning goals: by the end of this learning experience I would like to be able to demonstrate a convincing display of five magic tricks: the standard vanish, the tunnel vanish, the back clip vanish, the smart vanish, and the King Midas. Next, I would assess myself at the end of a certain interval. This assessment might be well served if it was defined in terms of the SOLO taxonomy.

How might a SOLO taxonomy look for a coin magician marking rubric? Lets take a stab at it. The SOLO has five levels of understanding and, according to Wikipedia, the five levels are defined follows.

1. Pre-structural: The task is not attacked appropriately; the student hasn’t really understood the point and uses too simple a way of going about it.
   
2. Uni-structural: The student’s response only focuses on one relevant aspect.
   
3. Multi-structural: The student’s response focuses on several relevant aspects but they are treated independently and additively. Assessment of this level is primarily quantitative.
   
4. Relational: The different aspects have become integrated into a coherent whole. This level is what is normally meant by an adequate understanding of some topic.
   
5. Extended abstract: The previous integrated whole may be conceptualized at a higher level of abstraction and generalized to a new topic or area.

Assume the following are potential statements from the perspective of a teacher of coin magic making an assessment on one of their students demonstrations of learned material:

1. Pre-structrual: The student simply holds the coin between his right hand pointer and index fingers, grasps the coin with the left hand and makes no effort to hide the palming of the coin in the right, then opens the left hand showing that the coin is not there, but clearly there is no illusion.
   
2. Uni-structural: The student demonstrates all of the steps of the required illusions poorly except for the palming aspect in each illusion. It is clear that the student has focused on only one aspect.
   
3. Multi-structural: The student demonstrates proficiency in each of the components of the illusions, but flow between each is poorly done, mechanical, and obvious.
   
4. Relational: The student demonstrates the required illusions in a convincing way, each step flows naturally to the next. The student shows promise.
   
5. Extended Abstract: The student demonstrates the required illusions in a natural, convincing way, but goes beyond the requirements of the assessment, narrating each step while they go through the demonstration, simultaneously discussing how each movement component could be utilized for other illusions, improved upon, or which ones are weaker.

To be good teacher to myself, by establishing learning goals in terms of the SOLO taxonomy, I am demonstrating constructive alignment. Good teaching gets students to use their higher level cognitive processing (evaluating, assessing, critiquing) instead of simply memorizing (like in surface learning). The goal is to make a surface learner behave like a deep learner by engaging with their material. To learn we must do the thing, whatever that might be.

This weeks progress

I established that each trick would be practiced 50 times over the course of a week. Completing the practice this week took dedication, and made my hands sore. Not only that, but I was forced to endure what seemed like endless dropping and retrieval of coins. This was certainly an exercise in patience. It would seem that self-assessment by utilizing the SOLO taxonomy in a way that I outlined above could be an exercise in futility, since I am biased of my own skills, and I am not an authority in the area of coin tricks. However, even if I cannot assess myself according to SOLO, I can employ the other aspects of deep learning simply by establishing learning goals in terms of SOLO, and keeping myself engaged with the learning task at hand.

Featured image by Krika99 at Flikr
https://www.flickr.com/photos/48499944@N03/14634917023

This post will serve as a learning plan for the Personal Learning Challenge that takes place throughout the EDCI 335 course.

I’ve always been interested in magic. A friend of mine can use magic to influence people in a very positive way. I noticed that his skills in magic resulted in a powerful social tool as an ice-breaker or a way to make friends. I have always admired his skills, but I don’t think they don’t come easy. At one point my friend showed me a card trick that was quite impressive. I asked him how long it took to learn, and he indicated that he stayed up all night learning the trick until his fingers had blisters. This type of dedication is quite admirable. However, a few questions arise in analyzing this technique of learning card tricks. If I wanted to learn a few card tricks of my own, would I need to stay up all night until my fingers had blisters? Would the same trick take me a lot longer to learn than my friend given that I have next to zero magic trick experience? How many times would I need to practice a magic trick skill until I could convincingly execute a non-trivial magic trick? Is there a difference between learning an abstract concept and learning a physical manipulation concept? It is important to consider the answers to questions like these when setting out to learn a new skill.

My aim over the next five weeks will be an attempt to learn a series of at least 5 coin tricks. I have zero coin trick experience aside from attempting to “palm” a coin, which means hiding a coin in your hand in such a way that one’s hand appears natural and relaxed to an outside observer. In learning these tricks, I will attempt to apply some basic learning principles based on concepts from the UVic EDCI 335 course, and from supplementary materials based on my own research.

Clarissa Sorenson-Unruh, in her post “Learning – The Neuroscience and the Neuromyths” indicates that three things occur when we learn something: encoding, consolidation, and retrieval [1]. How does learning a coin trick differ from that of learning an abstract concept like calculus for example. Learning coin tricks certainly involves encoding, consolidation, and retrieval, but it is much different from the kinds of learning done in a typical university setting. During most of my time at UVic, much of my effort has been spent learning mathematics, programming concepts, and algorithms. This involved primarily visual and semantic encoding, with very little tactile encoding [2]. Instead, the focus is on looking at the steps for a specific trick, reading or hearing a detailed explanation of each step, and then deliberately moving through the steps one at a time until these steps can be replicated without the crutch of the learning material in front of one’s eyes. Here we see the encoding portion which involves visual encoding by reading a book or watching an example video. Then, there is tactile encoding which involves manipulating a coin, feeling the weight of it, the shape of it, and moving it in such a way as to convincingly complete a trick and fool an observer. We then consolidate the memory of our slow and deliberate action, completing the coin trick by continuously practicing the movements step-by-step over and over until the trick becomes nearly instinct. The movements will become fast and smooth and almost unconscious. Years later, these tricks should presumably be replicated or learned again quickly through the process of retrieval assuming that the initial encoding and consolidation steps were successful.

One could argue that mathematics can be learned in the same way: write out the procedure to solve this one specific problem, manipulating the pencil in such a way as to make the exact same strokes and marks on the paper until such time as the movements of the pencil eventually result in the correct answer for the one specific problem every time. Tactile encoding might be successful for one problem, but issues arise when we are required to apply our knowledge dynamically. One change in value or small deviation in a math problem will render our purely tactile memory of how to solve a math problem totally irrelevant. This is where we see the largest difference between tactile and semantic encoding. While learning a coin trick and learning a calculus technique both involve visual and/or audio encoding, one relies primarily on tactile encoding, and one relies primarily on semantic encoding. However, both skills rely, at least in part, on a combination of all types of encoding. Tactile encoding is required to manipulate a pencil correctly, and there is some evidence to suggest that writing out the solution to a problem results in better learning outcomes [3], but semantic encoding certainly is the primary action that occurs in learning how to apply a general mathematical concept to a dynamic set of problems. Semantic encoding is required to understand how one small aspect of coin manipulation can be applied to many types of coin tricks, coin palming (hiding a coin convincingly in the palm of one’s hand) but simply reading a book over and over, writing out the steps for a coin trick ten-thousand times is no substitute for practicing that same trick with physical coin manipulation even one-hundred times. Certainly, tactile encoding is the primary mechanism for learning coin tricks. This means I will need to physically practice these tricks over and over until they become natural.

Based on this analysis of the types of encoding required for different skills, I will attempt to learn at least five non-trivial coin tricks in the coming four weeks using physical manipulation of the coins and not simply theoretical concepts of the tricks themselves.

I plan to learn and track the progress of my learning experience with the following strategy:

1. Research of the 5 coin tricks and a description of their outcome
2. Daily progress report on each of the tricks
3. A video documentation of the level of proficiency based some number of practice sessions.
4. A mirror used for immediate feedback and adjustment of technique or behavior for successful completion of the coin trick.
5. A comparison between my first attempts at learning each trick, and my final attempts.
6. A final analysis of my experiences of what worked and what did not.

Assessment of the success or failure of the learning of my new skill will be nearly immediate. If these skills are practiced in front of a mirror, or recorded on video and played back, I will be able to have an immediate indication about what went wrong, or where my skills can be improved. This principle is similar to B.F. Skinner’s approach in his theory of programmed learning. While not exactly programmed learning (since I’m not following a pre-designed program of step-by-step coin magic skills) I will still employ the principle of immediate feedback, which is what Skinner’s learning machine was based on [5]. In a sense, I am using Skinner’s behaviourism approach since the success of each practice session can be based on positive or negative feedback. If my trick is successful, the illusion will be convincing and natural looking. On the other hand, if the trick is unsuccessful, the illusion will not be an illusion at all, but rather an amateur attempt that fools nobody. I wouldn’t necessarily say that my learning plan is based on an Objectivist approach since these skills aren’t based on learning a body of knowledge, rather, they will be based on learning small aspects of a very large body of knowledge (a book of coin magic for example) and practicing these small aspects basically through rote-repetition.

Some preliminary research into coin magic books reveals one very promising option by author and magician J.B. Bobo who wrote “Modern Coin Magic” published in 1952 [7]. I will list the tricks themselves, and their effects, but not the underlying steps taken in order to achieve their effects. I will exclude the step-by-step trick method details since this blog is publicly accessible and the trick materials themselves are a purchased product:

This book was available for a reasonable price on Amazon, so I purchased a digital copy of it.

Browsing the book for coin tricks which range in difficulty from simple through difficult, I chose five tricks which I believe covers a wide range of trick difficulty levels, and that tricks looked interesting to me. The tricks that I will attempt to learn over the coming weeks are as follows (excerpts taken, in part, directly from “Modern Coin Magic”):

The Standard Vanish (Pg 21. J.B. Bobo’s “Modern Coin Magic”)
The coin rests near the ends of the two middle fingers of the right hand, Fig. 1. Right hand describes a counter clockwise movement, turning back upwards as the fingers curl inward and press the coin into the classic palm position where it is retained. This action takes place under the guise of supposedly placing the coin into the left hand, Fig. 2. The left hand closes as if it holds the coin. Look at and point to the left hand. Then snap the right fingers at the left hand. Open the left hand slowly and mysteriously. The coin is gone.

This trick appears to be on the easier end of the spectrum since it is just a “coin vanish” type which involves few steps. It is one of the first tricks one learns in Bobo’s book. From a beginners perspective, I rate this trick as elementary.

The Tunnel Vanish (Pg 24. J.B. Bobo’s “Modern Coin Magic”)
Hold the left hand palm downward and close it into a loose fist so only the thumb and forefinger touch. The right hand holds the coin horizontally between the forefinger and thumb – thumb is on top Fig. 1 … Left hand closes into a tighter fist and the right forefinger gives a final poke into the left fist. It appears that the coin was pushed into the left fist and then given a final poke with the right forefinger. The left hand is then turned palm up, opened, and shown empty. The coin has faded into nothingness…

This is another coin vanish, but there are more steps involved and appears to be slightly more difficult than the Standard Vanish. I rate this trick as potentially intermediate.

Through a Handkerchief (Pg 68 J.B. Bobo’s “Modern Coin Magic”)
A coin in caused to penetrate the fabric of a pocket handkerchief in a baffling manner…

This trick involves many more steps than either the Standard Vanish or the Tunnel Vanish. I would rate this as somewhere between intermediate and advanced difficulty.

Smart Coin Trick (Pg 83 J.B. Bobo’s “Modern Coin Magic”)
A borrowed half dollar is balanced on the tip of the right forefinger. The left hands forms a fist around it. The right hand is withdrawn and shown empty back and front, fingers wide apart. Right hand removes a handkerchief from the breast pocket, and holds it by one corner. Left hand is slowly opened, palm toward spectators, fingers wide apart. The coin has faded away!

This trick involves many more steps than Through a Handkerchief and involves coin and prop transfer to and from both hands. I am not a magician, so I will estimate the level of difficulty of this trick at advanced.

The Touch of Midas (Pg 90 J.B. Bobo’s “Modern Coin Magic”)
The conjurer shows his left hand empty and closes it. A spectator touches his left wrist. When the hand is opened it contains a coin! The coin is removed. This procedure is continued until four or five coins are produced.

Given my limited knowledge of the level of coin magic, I will rate this one (relative to the level of difficulty for beginners) as advanced, since it involves potentially impromptu palming or hiding multiple coins from observers.

Certainly, learning these tricks will be challenging, especially if the final result will produce a convincing illusion. I look forward to the challenge of learning the tricks themselves, and to documenting my learning experience within the context of the EDCI 335 course.

References

[1] https://clarissasorensenunruh.com/2019/04/20/5r-adult-learning-assignment-learning-the-neuroscience-and-the-neuromyths/

[2] http://www.human-memory.net/processes_encoding.html

[3] https://www.npr.org/2016/04/17/474525392/attention-students-put-your-laptops-away

[4] https://opentextbc.ca/teachinginadigitalage/part/chapter-2-the-nature-of-knowledge-and-the-implications-for-teaching/

[5] https://www.youtube.com/watch?v=jTH3ob1IRFo

[6] Featured image credit to Flikr user reynermedia
https://www.flickr.com/photos/89228431@N06/

[7] https://www.amazon.ca/Modern-Coin-Magic-J-Bobo/dp/0486242587/ref=tmm_pap_swatch_0?_encoding=UTF8&qid=&sr=