No coding required to start. Bring a topic you are curious about.
Algorithms now read culture on a scale no person can match. What you are shown on Netflix, Spotify, Audible, and Reddit is selected by systems designed to model taste and predict response. The same works are being used to train commercial AI models, largely without permission and with little public visibility. Our culture has been converted into data and turned over to machines, and the process is mostly hidden from us.
This course examines that process directly, by having you carry out a version of it. You will take a body of work you care about and represent it as data a computer can analyze, then assess both what the analysis reveals and what it leaves out.
You choose the subject: a genre of fiction, a musician's lyrics, a collection of artwork, an online community, or any cultural material you want to understand better. Most of these are normally encountered one item at a time. This course is concerned with what becomes visible across thousands at once, a full genre, a songwriter's complete catalog, a hundred thousand fan fiction posts, and with what is lost when computational analysis replaces a human reader.
The majority of the course is spent investigating a single question of your own. For example:
The central difficulty is deciding how to measure things, and that is where most of the thinking happens. To count something, you first have to define it. Is a song "happy" by virtue of its lyrics, its melody, or the comments beneath the video? Where does one genre end and another begin? Such decisions determine the results.
You will apply the same scrutiny to the systems that shape what you read and watch, and to published research. A lot of publications in machine learning and adjacent fields were produced by people who made these same decisions, usually without reporting them and sometimes in error. You will learn to identify the choices behind a research project, evaluate whether they hold, and document your own.
Coding is not required. An AI assistant writes the code, so you can begin on the first day. You will learn to read its output and verify it, since these tools are often wrong in ways that appear convincing, and the objective is to use them without being misled. Those who already program, or wish to learn, may work with the code directly and extend their projects accordingly. We will start with counting words and pixels and work our way up to the more complicated machine learning techniques that enabled modern AI systems.
We will have a mix of lectures, discussion, and collective project support. You will review examples of what is possible, select your question, and develop it into a project with input from your classmates and the instructor. Questions are expected, and no prior background is assumed. In the final weeks you will write up your findings as a short web page, combining visualizations and text to present both your results and the process behind them, and present the work to the group. You are welcome to get inspiration from data journalism projects like The Pudding (https://pudding.cool/2026/05/similes/) or the New York Times Upshot (https://www.nytimes.com/2026/02/18/upshot/moltbook-artificial-intelligence-ai.html), but it can take whatever form best conveys your discoveries!
I’m Lucian, a PhD candidate in Information Science. My research focuses on using machine learning to study how ideas spread in academic networks and how scientific discoveries influence and are influenced by popular culture. I am most interested in reading and watching anything science fiction, but I love learning about all types of cultural expression, and I’m looking forward to learning from everyone’s fascinating projects this term.
Want to join? Fill out the application form below!