Human-Machine Teaming Games

The Human-Machine Teaming (HMT) Games project builds games as research platforms for studying how humans and AI systems collaborate. The project has produced a suite of five games, each targeting a distinct facet of the teaming problem — coordination, communication, planning, delegation, and situational awareness — along with the Hierarchical Puppetry Interface, a control system that allows multiple humans and agents to collaboratively manage in-game entities. It is a collaboration between CMU and Christopher MacLellan’s group at Georgia Tech, supported by the Army Research Lab STRONG program.

Dice Adventure — an asymmetric cooperative game in which human and AI teammates hold different information and abilities, and must coordinate to progress.

Vertical Farm — a cellular-automata farming task whose core challenge involves situational awareness, diagnosis, planning, and coordination among multiple humans and agents.

The games are used in open competitions that serve as both evaluation and engagement. Competitors participate either as human players or by contributing an AI agent of their own design; humans and agents are matchmade into hybrid teams, with the highest-scoring human-AI team taking the prize. Dice Adventure anchored the inaugural competition at the IEEE Conference on Games, sponsored by IEEE with award funding, and the paper describing the game and its competition design received a best paper award at the Foundations of Digital Games conference.

Alongside the games, the project is developing a framework for mapping the dimensions of the HMT design space, clarifying how different task characteristics contribute to teaming complexity and performance.