Mechanix is an interactive game and tangible interface for building chain reactions and learning from examples.

The premise of the game is to arrange tangible pieces on an interactive display to guide a marble to a target. Using an image detection system, designs people created were automatically saved, enabling anyone to see and test example solutions from other players.


Mechanix started out as a class project at Stanford for Beyond Bits and Atoms, a course in the graduate school of education. I collaborated with my good friend Coram Bryant on the project for a couple years after we took the class.

The initial sketch for Mechanix from my notebook in 2010

The game consists of magnetic physical pieces that can be freely arranged on the screen. Each piece represents a type of "simple machine," such as a lever and wheel and axle.

Simple Machine Pieces

On the back of each piece is a fiducial marker, an image marker that can be used to uniquely identify each piece using a webcamera placed behind the display. One of the innovations of designing the Mechanix hardware was determining the combination of materials that would enable us to both see the pieces behind the screen, yet also enable the pieces to stick to the display. We used a combination of magnetic steel mesh, vellum paper (as a projection surface), and hard acrylic for giving the screen structure (we actually filed a provisional patent on this design).

Labelled simple machines

In the game, Rolly the marble wants to get back to his friends, and you're tasked with arranging the pieces to guide the marble back "home."

Kids had a ton of fun playing Mechanix! We shared the game at many different events, including Maker Faire Bay Area 2011, where we got an Editor's Choice award.

That moment when your design works!


From a research perspective, we were interested in exploring about how we could enable children to learn through building and testing other people's examples. With many tangible toolkits, it's hard to capture what people create in a way that's accessible to other people. The Mechanix design enables us to record people's saved designs automatically, and projecting the examples directly onto the display so you could test them out.

Reviewing examples with Mechanix

For more on education studies we did looking on children's uses of examples, check out the paper we wrote for IDC.

In our initial studies with children, we found that they often learned about pieces they hadn't used before or discovered interesting combinations to try when looking at other people's examples. However, they didn't necessarily choose to look at examples on their own:

This led us to experiment with different interfaces for reviewing examples, including prompting users asking them to compare their solution with another, and viewing a portfolio showcasing the examples you created.

Side-by-side comparison

A portfolio summarizing the examples you've created

We found these small changes to Mechanix inspired some surprising behavior:

For more on these reflection interfaces, check out our WIP paper presented at CHI.


Through our research with Mechanix, we saw that automatic documentation of examples exposed children to different approaches to combining simple machine pieces, enabed them to experiment with unfamiliar pieces, and served as a portfolio with which children can review their own designs. Additionally, the vertical display and multiple points of interaction afforded by the tangible components enabled children to engage in collaborative design.

Our work experimenting with reflection interfaces found that incorporating children’s reflections into subsequent interactions motivates learners and makes reflections meaningful. Multiple reflection interfaces help broaden the reflection experience for each user and support varied preferences across all users. However, the reflection modes should be appropriately matched with children’s design behavior to maximize learning potential. Finally, audio recordings can enhance learning by providing opportunities to communicate understanding, reinforce interface-specific vocabulary, and forefront misconceptions.