Social Graphs
and Interactions

The new way

Below is an overview of everything you need to know. I've built the course to embrace that AI changes everything about how we work with data. So there are two separate strands going through every lecture:

• Expanding your brain. I need to motivate you to actually learn something, and I need a way to test that you have actually learned something. So twice during the semester you will need to do an in-class, pen-and-paper test. Closed book, no devices. Nobody likes those tests. It's tough to face your actual capabilities. But it's the only way to see who you are without an all-powerful AI wingman by your side.
• Using AI to make teaching better. Everywhere else, use every tool you have. And it means we can raise the bar of what's possible: we'll be using real datasets you crawl yourself, interactive visualizations, simulations, complex tools. Things that there's no way I could have even imagined doing in 2020.
• Responsible AI driving: And there is also a brand new scientific skill that I'll try to push you to use all semester: validating what the machine tells you. An LLM that labels your data is a tool, and tools need to be calibrated, tested and checked.

Every exercise carries one of three modes

The weeks

I use a method called flipped classroom. The idea is that in class you work through a page of carefully designed material. Weeks 1–4 are the networks half, weeks 5–7 the language half, and in week 8 we combine everything.

I end every week with a Go nuts with your LLM exercise. It's a free-form, agent-powered analysis on a shared dataset that you should do in groups, then publish on your group's own site, with a weekly highlighted project. Weeks 9–13 are the project period.

Course overview

One table, every date that matters. We meet 13 Wednesdays from 09:00. In weeks 5 and 9 the day starts early with the test, 08:30–09:30 (rooms below), then class continues as usual.

And the practicalities:

Groups

The idea is that you should work in a 3-person group. Try to form one as early as possible during the first session(s). Then you work through the weekly pages in parallel, helping each other, discussing strategies, etc. The go-nuts posts are designed for groups (it's just more fun with friends to brainstorm and discuss alongside), as is the final project.

Groups smaller than three people are OK, but larger groups are not. The aim is to form a group in the first session, but if it takes a couple of lectures that's also OK.

There is a single iron rule: everyone must be able to solve everything. The tests are individual, on paper, and they do not care who in your group understood it.

The two tests

Twice during the semester you sit down with pen and paper, in class: closed book, no devices. Test 1 (Wed 30 Sep) covers weeks 1–4; Test 2 (Wed 4 Nov) covers weeks 5–8; each counts 25%.

Each test is ~60 minutes: 16 multiple-choice questions, and 4 tasks where you draw, calculate and explain. The tests evaluate transfer, not recall: reading figures, predicting outcomes, spotting flawed analyses, small hand calculations, and choosing the right method. Everything is drawn from the 🧠 Learn-mode material — and every week page ends with an "on the test" box telling you exactly what you should be able to do on paper. No surprises, no trivia: if you did the 🧠 exercises with your own head, you are prepared.

Both tests take place in building 303A, Auditoriums 42 and 43. For Test 1 you got an email telling you which of the two rooms is yours; for Test 2, go to the same room. I want you to be there at 8:10 at the latest. We use 8:10–8:30 to get everyone seated and set up with pen and paper, and the test itself runs 8:30–9:30, starting sharp. If you arrive later than 8:10 it will count as a score of zero. If you are ill or otherwise not able to attend, do let me know ahead of time. There will be one make-up test. Details on that will be made public later. But it will not be to your advantage to wait.

The final project

The overall idea is this. In your group,

• Pick an area that you are excited about
• Find a dataset about that domain (much more on this later)
• Use the tools you've learned in the class to explore the dataset and find something new and exciting about your dataset that nobody knew.
• Deliver two things: a short pdf (the methods section) that we will evaluate. And a short in-person presentation (the demo fair) where we can ask questions.
• Dates: The project runs over weeks 9–13, counts 50%, and the demo fair is on 2 Dec.

I'll provide more info on this part later in the class.

Grading

Test 1 (25%), Test 2 (25%), and the final project (50%) will be scored, and I'll provide each score as feedback so you can see how you're doing. For the final evaluation, I map the three scores holistically to a final grade on the Danish 7-point scale. More details to come.

The book

Michele Coscia, The Atlas for the Aspiring Network Scientist (2nd edition) — free PDF, with per-chapter exercises and solutions on the site. Weekly pages point to exact chapters. A few classic papers join it along the way.