Truth About Human Learning: A Stress-Free Guide to Mastering AI/ML

Table of contents

Introduction

Hey everyone! Being in an AI/ML PhD program is a massive undertaking. If you’ve been feeling like the Python code isn’t “clicking” or the Math formulas look like a foreign language, you are definitely not alone. Most of us start by trying to “absorb” the material by reading it over and over, but that often leads to burnout rather than mastery.

The good news? It’s usually not a “you” problem—it’s a strategy problem. By shifting our focus from passive learning to active learning, we can actually spend less time studying while retaining much more.

Here’s a breakdown of how to make your study sessions more effective and a lot less stressful.

Why “Hard” is Actually Good: The Science of Learning

Before we dive into the “how,” let’s look at the “why.” We often feel discouraged when we struggle to remember something, but research shows that the struggle is the signal that you are actually learning.

Below are two common traps and corresponding researches. More researches are attached at the bottom.

The Re-reading Trap (Roediger and Karpicke, 2006)

Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: taking memory tests improves long-term retention. Psychological science, 17(3), 249–255.

In a famous 2006 study by Roediger and Karpicke, researchers compared two groups of students:

• Group A (Repeated Study): They read the material four times.

• Group B (Single Study + Testing): They read it once and then took three practice tests.

The Result: While Group A felt more confident immediately after studying, they forgot almost everything a week later. Group B scored significantly higher on long-term retention. Taking a “quiz” (even if you fail it) forces your brain to build much stronger connections than simply re-reading.

The “Fluency” Trap (Kornell, 2009)

Kornell, N. (2009). Optimising learning using flashcards: Spacing is more effective than cramming. Applied Cognitive Psychology, 23(9), 1297–1317.

Research by Nate Kornell showed that students who cram often perform just as well as “spacers” on a test taken immediately after the session.

The Quantitative Results:

• 90% of participants performed better when they used spacing rather than cramming.
• On the final test, the spaced group scored significantly higher (54%) compared to the cramming group (42%), despite both groups spending the exact same amount of time studying.
• The “Fluency Trap”: Interestingly, 72% of students thought cramming was more effective for them, even though their own test scores proved it wasn’t.

Conclusion: The “Desirable Difficulty” Principle

Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In M. A. Gernsbacher, R. W. Pew, L. M. Hough, & J. R. Pomerantz (Eds.), Psychology and the real world: Essays illustrating fundamental contributions to society (pp. 56–64).

Psychologist Robert Bjork coined the term “Desirable Difficulty.” He found that when your brain has to work hard to retrieve a memory (that “tip of the tongue” feeling), it’s actually a biological “save” button.

If it feels easy, you’re likely just recognizing the information. If it feels hard, you’re actually encoding it. So, when you struggle to recall a Python function or an AI/ML concept, don’t be discouraged—that’s your brain literally building a stronger neural path!

Comparing Learning Strategies: What Works?

To help us move away from the “busy work” and toward actual growth, here is how common strategies compare. Notice how the “High-Efficiency” side is all about doing rather than just viewing.

Putting it Together: Your 5-Step AI/ML Workflow

Instead of doing these in isolation, try this integrated flow. Let’s use Decision Tree as our example:

1. Elaborative Interrogation (The Foundation)

As you read about Decision Trees, ask yourself: “Why do we use Entropy or Gini Impurity to split a node?” (Answer: To find the feature that narrows down the possibilities most effectively, creating “purer” groups of data).

2. The Feynman Technique (The Clarity Check)

Explain how a Decision Tree works to a “rubber duck” or a friend.

• The Analogy: “It’s like sorting a mountain of mixed laundry. You don’t just grab random items; you find the one rule that creates the ‘purest’ piles—like separating ‘Whites’ from ‘Colors’—because that reduces the chaos (entropy) the fastest.”

3. Interleaved Practice (The Integration)

Don’t just read the theory.

• Math: Manually calculate the Information Gain for a tiny dataset (e.g., 4 rows) on a piece of paper.

• Python: Immediately switch to your code editor and use DecisionTreeClassifier from Scikit-learn to see if the model’s first split matches your manual calculation. Switching between the “pen-and-paper math” and “library implementation” solidifies the link.

4. Active Recall (The Stress Test)

After an hour, cover your notes. Try to write down the Entropy formula $H(S) = -\sum p_i \log_2 p_i$ or sketch a small tree for a simple logic gate (like AND or OR) from memory. Don’t worry if you get the log base wrong—the struggle to remember is what creates the “save” point in your brain.

5. Spaced Repetition (The Long-term Lock)

Create an Anki card for the specific difference between “Information Gain” and “Gini Impurity.” Review it tomorrow, then 3 days later, then a week later.

Fortunately, I have created an Anki deck about AI/ML. DM me if you need a copy.

Researches

Cramming vs. Spaced Repetition (Cepeda et al., 2006)

Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological bulletin, 132(3), 354–380.

A massive meta-analysis by Cepeda and colleagues found that “spacing” out sessions resulted in a significant increase in test scores—sometimes by as much as 10% to 30%—compared to massed practice (cramming) for the same amount of total study time.

• The Takeaway: You can spend 10 hours cramming on Sunday, or 1 hour a day for 10 days. The 10-day group will remember the material for months, while the Sunday group will forget it by Tuesday.

Rote vs. Feynman (Chi et al., 1989 & Fiorella, 2013)

Chi, M. T. H., Bassok, M., Lewis, M. W., Reimann, P., & Glaser, R. (1989). Self-explanations: How students study and use examples in learning to solve problems. Cognitive Science, 13(2), 145–182. Fiorella, L., & Mayer, R. E. (2013). The relative benefits of learning by teaching and teaching expectancy. Contemporary Educational Psychology, 38(4), 281–288.

The Feynman Technique is effectively Self-Explanation. A classic study by Chi et al. (1989) found that “high-explainers” (students who explained the why of each step to themselves) solved twice as many problems as those who relied on rote memorization.

• Success Rate: ~86% for self-explainers vs. ~42% for rote learners.

• Learning by Teaching: Research by Fiorella & Mayer (2013) found that students who prepare to teach others have a median effect size of $d = 0.77$ (a “large” effect) in comprehension compared to those who just study for a test.

The “Desirable Difficulty” Principle: Psychologist Robert Bjork found that when your brain has to work hard to retrieve a memory (that “tip of the tongue” feeling), it’s performing a biological “save” button. The hardness is not a sign of failure—it’s the sound of neural connections getting stronger.

Final Thoughts

We’re all in this together! The goal isn’t to be a perfect coder overnight—it’s just to make our learning a little more intentional. Next time it feels “hard,” take a deep breath and remember: that’s just your brain getting stronger.