How I Prepare for Exams (and Actually Retain Things)

After 2+ years of exams in the AI bachelor, I've tried pretty much everything. Colour-coded notes, marathon study sessions, re-reading slides until my eyes glazed over. Most of it didn't work. Here's what did, and why.

Stop re-reading your notes

I'll start with the thing nobody wants to hear. Re-reading slides, highlighting, rewatching lectures: it all feels productive, but it barely moves the needle. Dunlosky et al. (2013) reviewed hundreds of studies and ranked re-reading and highlighting among the least effective study methods. The reason is that passive review creates a false sense of familiarity. You recognize the material, so you think you know it. But recognition is not recall. The exam doesn't ask "have you seen this before?" It asks you to produce answers from nothing.

Active recall

This is the single biggest change I made. Instead of passively going over material, I close my notes and try to reproduce what I just studied. Write it out, say it out loud, whatever. It's uncomfortable because you realize how much you don't know, but that's the point.

Karpicke and Roediger (2008) tested this: students who used active recall retained 80% of information after a week, compared to 36% for students who just re-read the material. That's not a small difference.

What this looks like in practice:

• Read a section, close the book, write down everything you can remember, then check what you missed.
• Use Anki or physical flashcards.
• Do practice problems before you look at solutions. Struggle with them first.
• Try the "blank page method": take an empty page and recreate the key ideas from memory.

Spaced repetition

Cramming the night before is tempting because it feels like you're covering a lot of ground. You are, but you'll forget most of it within a week. Spaced repetition is the opposite approach: review material at increasing intervals over days and weeks, so it actually sticks.

A meta-analysis of 317 experiments (Cepeda et al., 2006) found that spaced practice can double or even triple long-term retention compared to cramming. It's not a new idea either; psychologists have been studying the spacing effect for over a century.

For a 3-week study period, something like this works:

Day	What to review
Day 1	Learn new material
Day 2	Review Day 1 material
Day 4	Review Day 1 material again
Day 8	Review Day 1 material again
Day 15	Final review before exam

Anki automates the scheduling for you. Create cards early (3-4 weeks before the exam, not 3-4 days) and let the algorithm do the rest.

Interleaving

This one is counterintuitive. Instead of studying one topic in a long block and then moving to the next, you mix topics within a single session. So instead of doing 20 linear algebra problems and then 20 calculus problems, you alternate: 5 LA, 5 calc, 5 LA, 5 calc.

Rohrer and Taylor (2007) found that interleaving improved test scores by up to 43% compared to blocked practice. It feels harder while you're doing it, which is exactly why it works. Your brain has to keep distinguishing between different types of problems, and that effort makes the learning stick.

Ask "why?" more

This is sometimes called elaborative interrogation, which is a fancy name for a simple habit. When you're studying a concept, don't just memorize the fact. Ask yourself why it's true and how it connects to other things you know.

For example: don't just memorize that gradient descent updates weights by moving in the direction of the negative gradient. Ask yourself why the negative gradient direction reduces the loss. What would happen if you moved in a different direction? What if the learning rate is too large?

This kind of questioning forces you to actually understand things instead of pattern-matching.

Draw things

Combining text with visuals (diagrams, sketches, concept maps) helps because you're encoding the same information in two different ways. If you can't recall the text, you might still remember the diagram, and vice versa.

For AI courses specifically: draw out neural network architectures, sketch decision boundaries, make concept maps that link related theorems together. It doesn't have to look good. The act of making it is what matters.

Teach it to someone

If you can explain a concept clearly to someone who doesn't get it, you actually understand it. If you stumble or wave your hands vaguely, you've found a gap. This works great in study groups: take turns being the "teacher" for different topics.

There's research showing that even the expectation of having to teach something changes how your brain processes it. You organize the material more carefully when you think you'll need to explain it.

Pomodoro (25 on, 5 off)

I used to sit down for 3-hour blocks and wonder why I couldn't focus past the first hour. Now I do 25 minutes of focused work, then a 5-minute break, and take a longer break after four rounds. It's nothing revolutionary, but it works. My concentration lasts longer and I actually get more done in less time.

Sleep, exercise, and your phone

Three things that have nothing to do with study technique but affect everything:

Sleep: Your brain consolidates memories while you sleep. Pulling an all-nighter before an exam is actively making things worse. A meta-analysis in Psychological Bulletin found that sleep deprivation can cut cognitive performance by 25-50%. Just go to bed.

Exercise: Even a 20-minute walk before studying helps with focus. I don't fully understand the neuroscience here (something about BDNF), but it works and it's free.

Your phone: Put it in another room. I'm serious. What people call "multitasking" is actually rapid task-switching, and according to APA research it can reduce how much you learn by up to 40%. Every time you check a notification, you lose focus and it takes minutes to get it back.

Practice problems

For math and programming courses, doing practice problems matters more than anything else on this list. Complete all the assigned problem sets, even the ungraded ones. Work through past exams under timed conditions. And actually try each problem before looking at the solution. Struggling with a problem you can't solve teaches you more than reading a solution you didn't attempt.

Reflecting on what worked

After a study session, I spend a few minutes thinking about what I actually learned, what was still confusing, and what I should do differently next time. This sounds like journaling, and I guess it is, but it's helped me stop wasting time on methods that don't work for me.

On exam day

• Sleep well the night before. Cramming doesn't help much for technical courses.
• Get there early. Do a quick review of key formulas.
• Read all the questions before you start. Do the ones you're confident about first.
• Show your work even if you're unsure. Partial credit adds up.
• If you're stuck, move on. Come back with fresh eyes later.