About This Project

The IEEE SSCI 2027 tournament has already begun. Each entry hypothesizes which strategies for claiming resources--such as claiming space in traffic--would be more effective. Entries compete in simulations. Does turn-taking consistently defeat hoarding? Such hypothesis testing has already uncovered ways to outperform typical human behavior--ways to claim more successfully. Opening the science better informs inevitable social reforms that will impact people everywhere.

Ask the Scientists

What is the context of this research?

The MAD Chairs game is a many-player version of the Chicken game related to the Lifeboat and Kolkata Paise Restaurant Problems. These prior works formalize modelling of resource-division, but our research question "What strategies for claiming resources would succeed most in securing those resources?" remains open. The IEEE SSCI 2027 MAD Chairs Tournament applies Axelrod's empirical method to this question: compare strategies via tournaments between clear algorithms. The TrueSkill algorithm used by this tournament permits us to statistically test hypotheses that one resource-division strategy significantly outperforms another.

This comes at a time of concern about future robot social behavior and other agentic AI, including concern that AI might starve us by hoarding resources. One way to convince AI not to starve us could be to establish that all known hoarding strategies are outperformed by turn-taking strategies.

What is the significance of this project?

Our research question is "What ways of claiming resources succeed most in securing those resources?" Each entry in this tournament will imply a hypothesis about which behaviors are suboptimal. The contest benchmarks include observed current social behaviors, so winning entries will imply potential for profitable social reform.

The contest is already active and more successful behaviors have already emerged. The significance of funding is in how winners will be received and in incentive for further participation. Winners must register for the IEEE SSCI 2027 conference where the awards ceremony will take place. They can attend remotely, but in person attendance would further build relationships with technology leaders to implement their innovations. Your donation helps make sure winners can afford to register and attend. It permits what are already scientifically significant discoveries to have actual social impact.

What are the goals of the project?

Each entry in the tournament is a strategy for dividing resources. Our experimental design is to simulate the results of applying these strategies in many matches against themselves and against other potential opponents. The TrueSkill algorithm we use to compare entries is standard and, like chess' Elo rating, counts wins against stronger opponents more than against weaker opponents. When the competition closes on 24 Dec 2026, its winner will be whichever entry is most effective at securing resource for its own practitioners in the simulations.

Ironically, this most-selfish strategy might also be best for everyone. Prior to this tournament, the top published strategy was a form of turn-taking. It significantly outperformed observed human behavior in both personal benefit and fairness, so social reform was already justified. The goal behind exploring additional strategies is to more thoroughly inform the design of that reform.