Ten simple rules for teaching data science

Imagine teaching data science is like teaching someone how to bake a complex cake. Most teachers might start by lecturing for three weeks on the chemical properties of flour, the history of ovens, and the physics of heat before letting anyone touch an ingredient. This paper argues that's a terrible idea. Instead, you should let them taste the cake immediately, then teach them the science as they crave more.

Here are the Ten Simple Rules for teaching data science, translated into everyday language with some tasty analogies:

1. Let Them Eat Cake (Start with Action)

Don't wait until the end of the semester to show students what data science actually does. In the very first class, have them load some data, clean it up a little, and make a picture (visualization).

• The Analogy: It's like giving a new driver the keys and letting them sit in the car before explaining how the transmission works. Once they feel the engine roar, they'll be begging you to explain the gears. If you start with boring technical details (like "object data types"), they'll tune out. Hook them with the fun stuff first; the boring technical details will make sense later when they need them to fix a problem.

2. The "Live Cooking" Show (Participatory Live Coding)

Don't just show students a finished slide with code on it. Type the code out in front of them while you talk through your thought process.

• The Analogy: Think of a chef on TV. If the chef just shows a photo of the finished dish, you learn nothing. But if the chef chops, burns a garlic clove, curses, fixes it, and explains why they are stirring the pot, you learn how to cook. When you make a mistake while typing, it's a good thing! It shows students that experts mess up too, and it teaches them how to fix errors (debugging) in real-time.

3. Drill, Baby, Drill (Lots of Practice + Instant Feedback)

Give students a massive amount of practice problems, way more than you think they need. But here's the catch: they need feedback immediately, not a week later when the grade comes back.

• The Analogy: Learning to juggle. You can't learn by watching one video. You have to drop the balls a hundred times. If you drop a ball and someone yells "Too high!" five minutes later, you've already forgotten what you were doing. Use automated tools (like software tests) that act like a coach standing right next to you, instantly saying, "Hey, you missed a comma," so you can fix it while the lesson is still fresh.

4. Use Toy Trains (Start with Small, Simple Data)

When introducing a new, scary tool or algorithm, don't throw a massive, messy real-world dataset at them. Use a tiny, "toy" dataset that fits in your head.

• The Analogy: If you want to teach someone how a car engine works, you don't hand them a whole truck engine with 500 parts. You give them a toy car engine with three gears. They can see every piece move. Once they understand the toy engine, then you give them the real truck. Small datasets let students track every step of the math without getting lost in the noise.

5. Real Food, But Easy to Chew (Real Data, Accessible Context)

Once they understand the toy engine, switch to real data. But make sure the data is about things everyone understands, like countries, populations, or languages.

• The Analogy: Imagine teaching a cooking class. If you use "deep-sea squid" as your main ingredient, students who have never seen the ocean will spend all their time wondering what a squid is, rather than learning how to cook it. Use ingredients everyone knows, like potatoes or carrots. The goal is to teach the cooking technique, not the biology of the vegetable. Use data about things people live with every day so they don't get overwhelmed by the subject matter.

6. The "Why" Behind the Weirdness (Cultural & Historical Context)

Explain why software tools are designed the way they are, even if they seem weird or annoying.

• The Analogy: Imagine you are driving a car with a gear stick that feels awkward. If you just say "deal with it," you'll get frustrated. But if you explain, "This car was built in the 1950s for a specific type of road, and that's why the stick is there," it makes more sense. Data tools are built by humans with history. Explaining that a weird symbol exists because of an old keyboard or a statistician's preference helps students stop hating the tool and start understanding it.

7. A Safe Playground (Inclusive Community)

Create a classroom where students feel safe making mistakes. If they are afraid of looking stupid, they won't ask questions, and they won't learn.

• The Analogy: Learning is like learning to dance. If you are terrified that everyone will laugh if you step on a toe, you'll stand frozen in the corner. You need a "no-judgment zone" where stepping on a toe is just part of the process. Set clear rules against bullying and harassment so everyone feels safe enough to try, fail, and try again.

8. The Checklist for Chefs (Peer Learning with Checklists)

When students review each other's work, don't just say "grade this." Give them a checklist.

• The Analogy: Think of a pilot before a flight. They don't just "hope" the plane is ready; they go through a checklist: Wheels down? Fuel full? Doors locked? When students review a friend's code, a checklist stops them from missing the small, critical details. It turns a vague "this looks bad" into specific, helpful feedback like "You forgot to check the data types."

9. The Band Practice (Collaboration)

Data science is rarely a solo act; it's a team sport. Teach students how to work together, not just how to code.

• The Analogy: You can be the best guitarist in the world, but if you can't listen to the drummer or share the stage, the band will fail. Teach them the "social tools" of collaboration: how to give feedback without being mean, how to use shared folders (like Git), and how to handle arguments when they happen. Practice this in small groups before throwing them into a big project.

10. The Whole Cake (Do a Full Project)

Finally, have students build a complete data product from start to finish.

• The Analogy: Most classes teach you how to mix batter, then separately how to bake, then separately how to frost. But in the real world, you have to do the whole thing. A final project forces students to take a messy, real-world problem, clean it, analyze it, and present it. It's the difference between practicing scales on a piano and playing a full song. It's messy, it's hard, but it's the only way to really learn the job.

The Bottom Line:
Teaching data science isn't about dumping facts on students. It's about getting them to do the work immediately, giving them safe spaces to fail, using simple examples to build big ideas, and eventually letting them cook the whole meal themselves.