Machine Learning for Design
Instructor: David Chuan-En Lin
Learning goals
By the end of the course, you will be able to:
- Use generative models and AI agents for design research, ideation and prototyping, and judge when they help compared with working without them.
- Fine-tune an image model on your own designs, and choose which designs and captions go into its training data.
- Explore and steer a model you trained, for example with concept sliders.
- Build AI tools for design work with coding agents, and decide which tasks to hand to an agent.
- Evaluate whether designers can direct, inspect and revise what an AI tool makes.
Schedule
| Week | Studio | Readings | Due |
|---|---|---|---|
| Part 1: Designing with AI | |||
| 1 | AI in the Design Process | ||
| 2 | Design Research with AI | ||
| 3 | Prototyping with Generative Models | ||
| Part 2: Appropriating AI | |||
| 4 | Training Data | ||
| 5 | Fine-Tuning and Steering | ||
| 6 | Building Tools with Coding Agents | ||
| Part 3: Agentic Design | |||
| 7 | Agents That Design | ||
| 8 | Simulated Participants | ||
| 9 | Evaluating Design Tools | ||
| 10 | Final Bake-Off | ||
| Finals | |||
Grading
| Part | Weight |
|---|---|
| Reading posts | 15% |
| Workshop hand-ins (weeks 1-6) | 20% |
| Project | 50% |
| Position essay | 10% |
| Participation | 5% |
Before each studio (starting with week 2), you will write a one-paragraph reading post about the required readings. The studios in weeks 1-6 each end with a short hand-in, due before the next studio. The course covers the cost of model training and API use.
Project
In the project, your team of 3 or 4 works as a small design consultancy for an organization you can reach (e.g., a campus design office, a local studio or a startup). You pick one design activity where the organization's designers have to make a choice (e.g., choosing between form concepts for a new product), and build an AI tool for it with any model or assistant. At the final bake-off, those designers and another team try your tool on a real task and rate whether they could direct it, inspect it and revise what it made.
| Milestone | Due | Weight | What to do |
|---|---|---|---|
| Team and partner | Week 2 | Form a team of 3 or 4 and find an organization to work with. | |
| Partner interviews | Week 3 | Interview 2-3 of the organization's designers, pick the design activity, and write 1-2 pages on what a better result would look like for those designers and how you will measure it. | |
| Midpoint | Week 5 | 10% | Give a 7-minute presentation of your proposal, with what you learned from the interviews and notes on related tools and papers. |
| Working prototype | Week 7 | Another team tries to break your tool in the studio. | |
| Partner tests | Weeks 8-9 | Test your tool with the organization's designers on the real task, and revise it. | |
| Bake-off and final report | Week 10 | 40% | Present at the bake-off, and hand in a 4-6 page report, a 2-3 minute video, the tool and its code, and a log of how you used AI. |
| Position essay | Finals week | 10% | On your own, write about 1,000 words on whether and how AI can design, using evidence from your own work in the course. |