I am reminded of a game we played in elementary school:

There are 100 pieces of candy in a jar, and 20 students. Each student gets to vote "share" or "selfish". If EVERYONE votes to share, the candy is split evenly. If ONE person votes "selfish", they get all the candy. If MORE than one person votes "selfish", no one gets candy, and the experiment is repeated until either the candy is distributed, or X iterations (3-5 seems normal) have passed.

Before each iteration, the students are allowed to discuss strategy. The solution, of course, is for a single trustworthy person to make a binding commitment to vote "selfish" and then evenly distribute the candy. By pre-commiting to vote "selfish", everyone else knows that voting "selfish" themselves will result in no candy - unlike a commitment to have everybody share.

I've always considered it a decent test of group rationality and social skills whether or not this consensus actually gets reached before the final iteration. I've seen groups that hit on this, had a single iteration with a few outliers testing it just to be sure that the person would really vote "selfish" like they said, and then implementing the strategy. I've seen others where 10-20% of the audience simply would not believe the person who made the pre-commitment, and so there was never a successful iteration