Evaluating Physicalizations for Augmenting Team Dynamics using a Tabletop Swarm UI

with Elyse Chase and So Yeon Park. Completed for Stanford’s CS 376 Introduction to HCI Research course in Fall 2018.

Abstract

Visual feedback of group conversation metrics such as speaking time has been explored as a means of balancing meeting participation and fostering effective teamwork. Recent advances in micro-robotics and Swarm User Interfaces have made real-time "physicalization" of this data feasible. As interactions with physical robotic systems have been shown to be more engaging and legible in some cases, we hypothesized that physicalizations of group metrics could be more effective at improving team dynamics than visualizations. To test this hypothesis, we developed a physicalization of group participation using the Zooids Swarm UI platform and performed a user study with 6 teams that used both our custom physicalization and an equivalent projected visualization for a series of brainstorming tasks. Results suggest that physicalization is more effective at balancing participation when controlling for between-group differences (p=0.02) and also results in higher perceived effectiveness among users.

Implementation

We used the Zooids swarm robot research platform for this project.

To relay relative participation information, we used a resource allocation metaphor. The table around which the group was seated always had a fixed number of "agents" (Zooids or projected circles), which we equally distributed among participants when conversation is balanced. If one person is dominating the conversation, they attract agents from the others proportionally according to participation.

To measure participation, each team member wore a microphone (Sparkfun Sound Detector, SEN-12642) to detect speaking time, connected to a central Arduino beneath the table. Speaking information packets were sent via Serial to the main program at 10 Hz, simply logging whether or not each member spoke in the time since the previous packet. In a main Processing sketch, a moving average calculated the relative speaking time of each person over a given time window.

A custom Processing application using the ZooidManager API converted this information into the spatial coordinates of all 20 agents, and either (a) displayed them in the form of colored circles on the tabletop via an overhead projector or (b) forwarded the positions to the Zooids which then positioned themselves on the table.

Technical Paper

Elyse D Z Chase, Eric Gonzalez, and So Yeon Park. 2018. Evaluating Physicalizations for Augmenting Team Dynamics using a Tabletop Swarm UI.