This is the result of session 6 in my course on Social Robot Design (2025/2026). The content of this session was mainly about designing a design tool to prototype the behavior of the robot, and also evaluate it on the robot itself.
Assignment: Design Behaviour Design Tools
Question to be answered
What can help you as a designer prototype (high level) robot behaviour? Especially tools that do not rely on a Wizard of Oz approach are welcome — or think of interesting hybrids. What could help you prototype and test?
- Flowchart, cognitive walkthrough, cards, board, software (ARC) — recipes
- Given that your phone/laptop have so many capabilities already — how to connect/interface this to motion and embodiment?
- How do you make ChatGPT-4o move a servo?
- Can you prompt ChatGPT to converse in a certain tone of voice, with predefined (bound) knowledge?
Project Notes
This idea started with a huge brainstorm essentially. At this point we didn’t have a direction yet.
Technical Plan
- Use borrowed Bluetooth Xbox controller for steering/moving the wheels
- Use the MIRO app to move the robot’s additional features (ears, eyes, etc.)
- Puppeteer the robot expression/scenario
- Decide when and why to move: wheels, eyes, ears, tail, head up/down, head left/right
- Other: sounds, lights, …
Design Tool Ideas
MakeYourOwnStoryboard
A storyboarding tool tailored to robot behaviour design.
Cards Against Humanities — Deck of Prompt Cards
Goal: Finding out robot hardware capabilities and exploring scenarios for robot behaviour.
CardsAgainstRobots
A game you play with multiple people where the goal is to find out the robot’s capabilities and explore topics related to the robot’s “job” (scenarios).
The robot in question might not have the capability to do what you want it to do, or functionality might not be appropriate to the use case. This step ensures that these are aligned. You also explore the different scenarios (and thus its behaviour) through using different types of cards: emotion cards, actuation cards, and wildcards.
In each step, the players pick one card of each to play out the scenario/behaviour. The first and second cards are randomly selected from the corresponding card piles. When the cards have been selected, the humans act out the scenario and puppeteer the robot and the ways the scenario could play out with different robot interactions — either through physical puppeteering or through the use of a controller.
The tool is deliberately built around one HRI idea: a robot action only means something once it’s paired with how a human is expected to read it. This is why every round forces you to draw an emotion card and an actuation card together. This makes it impossible to design “move ears” on its own, only “move ears because curious” or “move ears because startled”. A generic flowchart tool would let you wire up “move ears” as a transition without asking what that movement is supposed to communicate. The wildcard pile does something similar for graceful degradation as it throws in disruptions (an interruption, a sudden mood change) so the group has to also design how the robot’s behaviour recovers or adapts, not just the expected scenario.
Scenario Card Examples
| Human Action | Robot Action | Wildcards |
|---|---|---|
| Sits at desk frustrated | Go to desk | Tension |
| Mumbling equations | “That’s wrong” | Ensue crying |
| “Finally done!” | Happy move | … |
Card Types
| Emotion | Actuation | Wildcards |
|---|---|---|
| Happy | Move ears | |
| Sad | Lower head | |
| Annoyed | Move in circles |
Evaluation
We tested the deck directly on the MIRO-E robot in a short session where we were drawing cards and puppeteering (Controller connected to laptop through ROS). This was essentially a wizard-of-oz scenario, so it functioned as a brief functional prototype test of the tool.
Across several rounds, the deck did what it was supposed to do: drawing a human-action card, an actuation card, and a wildcard reliably produced a short, playable “play” of interaction within seconds, and the group could immediately act it out and see if it felt right.
Overall the deck worked well as a fast, easy way to get the whole team designing robot behaviour together, including teammates with no hands-on robot control experience, and the “student at a desk” scenario we generated with it fed directly into our group’s chosen direction. Its main weakness is that it designs which behaviour happens but not how it’s executed (speed, amplitude, timing), which may have mattered just as much as the choice of actuation itself. This matches findings from HRI research showing that varying only the speed and amplitude of an otherwise identical robot gesture significantly changes how people rate that gesture, without changing the gesture itself 1. Our card deck currently has no way to specify these execution parameters.
This was done with:
- Liz van Ginderen (s27349745)
- Anna Hornman (s3056600)
- Oyindrila Sen Gupta (s3697762)
- Sarah Mans (s2306379)
Deshmukh, A., Craenen, B., Foster, M. E., & Vinciarelli, A. (2018). Shaping robot gestures to shape users’ perception: The effect of amplitude and speed on Godspeed ratings. In Proceedings of the 6th International Conference on Human-Agent Interaction (pp. 293–300). ACM. https://doi.org/10.1145/3284432.3284445 ↩︎