C.A.N.V.A.S. is a speculative design project that seeks to explore personalities and functions of A.I. helper robots. For this project we also explored and critiqued roles for AI in politics. The C.A.N.V.A.S. Bot finds work on Earth after a local government organization promoting voter participation. C.A.N.V.A.S. Bot is the opposite of social media bots on Facebook and Twitter that pretend to be real people and are used in mass disinformation campaigns and political strategizing. People have grown suspicious of AI in this context, rightly so. How do their feelings change when the robot is made physical, something they can see and touch and interact with? What if the robot insists on remaining impartial and only providing logistical information?

C.A.N.V.A.S. from Martin Bernard on Vimeo.

In our scenario, the robots are from the planet Meon 3886 crash land on Earth after being discarded into space by the planet’s residents. The creatures on Meon 3886 have very short memories due to their short days, which last only 13 of Earth’s minutes. Furthermore the inhabitants live in very small, isolated communities, of only 150 individuals. Each of these communities on Meon 3886 has one of these little robots, whose job it is to record information and keep a public record that supplements the short memories of the aliens. The small communities on the planet all use the same robot, so that they can pool information, and every so often the robots are improved upon and the old versions are launched into space.

Coming from a planet so different from Earth, the robot has some trouble interacting with humans. This gave us an opportunity to explore ways in which AI helpers can be transparent in their interactions with human to allow for maximum understanding.

For this project we used the Delft AI toolkit, a system created by Phil van Allen that allows designers to quickly program behavior for a small robot that has several different AI capabilities. The programming is done in Unity, using a system of nodes that facilitates quick prototyping. The toolkit is open source, and can be found here: https://github.com/pvanallen/delft-toolkit-v2,, along with more information on the project.

The process for character development that we used for the project was designed to help us create strange AI personalities that resist anthropomorphic and skeumorphic default tendencies. First we imagined a random, made up, alien planet and civilization. Next we thought of robots that this alien culture might try to create. Finally we placed those robots in a strange environment, attempting to help the people of Earth. This process was helpful for trying to think of strange new ways that AI can communicate and interact with users.

One of my largest take aways from this project is how important yet difficult programming and developing a complex system of behaviors can be, and how difficult it is to think outside cliched tropes for robot personalities. Its important that we can do this because AI systems can be very helpful to us if we can design them in ways that are true to AI, that find interesting new methods for communication and collaboration.

This project explores the design of AI systems that help people with difficult things. Imagine how a small robot might assist people who have a disability, or are working in constrained situations (e.g. completely dark, where movement is restricted, etc.), or in life and death tasks such as working in space, or some other difficult circumstance. As a way to explore new approaches for AI and get past cliched design strategies, students must design the robot as if it already had an alien way of communicating and understanding the world that it was taught by its native civilization. Because it has this background, it communicates and behaves in odd ways that may seem strange to humans. Creating this backstory is part of the assignment. It is an intentionally strange, science fiction UX persona.

Delft AI Toolkit & Robot, Unity, Adobe Illustrator, Adobe Premiere Pro, Lumix G7

Class taught by Phil van Vallen

Project In collaboration with: Maya Friedman