An Interactive Workshop on Synthetic Mobility Data Generation

Real mobility data tells you where people actually go, and when. It is some of the most valuable material urban research has, and some of the hardest to share. Getting hold of it takes a data-sharing agreement, usually an ethics review, and months; even then you generally cannot pass it on. So a lot of work falls back on small self-reported surveys, or does not happen at all.

Synthetic mobility data is one attempt at a way round that: instead of releasing records of real journeys, you train a model on them and generate days that belong to nobody. It is a young idea, and it is not settled. How faithful the output really is, and how much of it could safely be published, are both open questions.

This is an interactive workshop on exactly those questions. Over an afternoon you will make some of this data, try to break it, and help challenge it.

Hosted by the UCL Centre for Advanced Spatial Analysis (CASA) and the PHOTO Research Team. A full programme will be circulated to everyone registered closer to the date.

Two ways to attend. Come in person for the whole afternoon, 13:00 to 17:00 at venue, which is the only way to do the hands-on parts and the discussion. Or join the 14:00 to 15:00 talk online. Choose when you book; online attendees are sent a joining link by email.

What you will do

• Invent someone's day. Paint one ordinary day — what a person does, and roughly where. Every imagined day goes up on the screen at the front as it arrives.
• Meet the real ones. Real observed days, put beside your inventions on the same screen, so the comparison is yours to draw rather than ours to assert.

• Generate one. Try the methods yourself. Take the approaches from the talk, run them on your own laptop, and compare what each one produces.
• Take the test. Real days and generated ones side by side, with no clues. Can you tell? Can the room? Can our own measures? We find out together.
• Make a ruling. Decide what should be publishable, and at how much detail. Your table's position goes into this project's real release documentation as a recorded outside opinion.

Practical

• Please bring a laptop. You will be drawing, generating and voting on your own machine, so a laptop is needed for most of the afternoon.
• No coding, no preparation, no prior knowledge. Everything runs in a browser, and every step is something you click.
• Free. Four hours with a break in the middle, tea and coffee provided, and it runs to time.

Info: Lunch is provided before we start. Come early if you can: we will help you get set up while you eat, which saves everyone time at one o'clock.

Info: You invent one ordinary day in someone's life: what a person does, and roughly where. We then look at real observed days alongside the ones you have made.

Info: This talk starts from how synthetic mobility data is actually produced. It starts from the three decisions any such model has to make: who a person is, what they do next and for how long, and where it happens. It then compares the main families of method in use today, from replaying real days to learning a representation of them, and reviews what is currently known about where each of them falls short. This hour is also streamed, and can be booked on its own.

Info: A blind test first: real days and generated ones side by side, and you say which is which. We reveal the answers together and compare your judgement with what our own measures catch. Then you go back to the generator, change its settings, and see what your changes do to the output.

Event venue & nearby stays

Maple House 149 Tottenham Court Road London W1T 7NF, Room 216 Maple House 149 Tottenham Court Road Lon, London, United Kingdom

Concerts, fests, parties, meetups — all the happenings, one place.

Get the app →

Discover more events by tag