User Study Plan

Goals

Try the system with a few people and collect feedback
One person at a time, iterative process
Asking people to modify one of our examples
We don't support regression yet, so pick example applications appropriately.
Target the week of March 7 to 11th.
Questions to ask: do you think this worked as well for you as for the author of this example?
What would make people look at the code and how do they know what's possible to change? Do we want a task that forces people to change the code (e.g. change the classifier)? What about a task that offers people multiple examples and asks them to figure out which one to use?
How do we help people know what to tune in the code and how to do it? (Since we don't have parameter tuning in the GUI.)
Need feedback on how to help people modify examples, not just on how the UI works.

Pre-Requisites (Organizers)

Example applications (e.g. video game controller). Generate mock keyboard events as control commands.
Using our laptops? Or making people install Xcode?
Arduino boards + sensors:

Pre-Requisites (Participants)

Workshop Steps

Ask about recording the session / screen / discussions.
Ask participants about their background with programming, circuits, and machine learning.
Explain what the project is, what machine learning is, at a high level.
Install our software (if participants are using their own computers)
Get everything running
Load Arduino code (we'll do for them)
Build circuit (we'll do for them)
Run our example, with the game as output
Show them how to modify training data.
Explain qualities of training data (e.g. what happens when you add more examples. temporal vs. static classification. null rejection).
Change something specific about our example (i.e. new training samples)
Calibration: e.g. use a training set from a different accelerometer
Overview of other examples
Do something more open-ended
Show people how to edit code / where to find documentation
Closing discussion / debrief.

Thoughts

Provide feedback