Highlights

• MazeWorld Testbed is a web-based Unity driven game that looks to gather quantifiable data about team interactions by tasking teams with collecting coins placed throughout a maze
• Challenge is to develop teamwork metrics based on observable events within the game

Project in Review

In recent years, many companies have started incorporating virtual assistants into their products such as Siri, Alexa, and self-driving cars. These features are added in order to increase the efficiency of the task the consumer is trying to complete.

Human-Agent Teams (HATs) are groupings of human and artificially intelligent software that work jointly to accomplish a given task. These combinations of humans and virtual assistances collaborate closely together in pursuit of common goals.. Challenges in HAT include quantifying the interactions between teammates, analyzing individual actions to determine the value and the need for both task skills and team skills.

The Mazeworld Study looks to gather data about the way that teams interact so that HATs can be developed to complete given tasks with greater efficiency. Participants are grouped together in teams of three to five and tasked with collecting coins inside a maze. Within the team, there are three different tasks that a participant could have tactical, explorer, and collector. The tactical position has a bird’s eye view of the maze and can direct their teammates to rooms that contain coins. Explorer has the ability to open doors, see inactivated coins, and activate coins. Finally, the rest of the team members are collectors who pick up all the activated coins. Throughout the hour-long session, participants will run three trials of this game in between these trials they will each be surveyed using the NASA Task Load Index (TLX) questionnaire. These questions are designed to understand how a participant feels about the physical and mental demands of the task. Additionally, there are also questions about team compatibility and strategy.

With the data gathered from the trials and the surveys, analysis can be done to explore what traits of the team increase performance and collaboration.

Funding

Game2Work, Iowa State University

Research Team

Dr. Michael Dorneich (PI), Dr. Stephen Gilbert, Elizabeth Cavannah, Yvonne Farad, Joseph Rozell, & Jonathan Segal