This market resolves once someone performs the following experiement, or provides proof that it's been performed in the past.

A random sample of pairs of identical twins is collected. Some general biases are acceptible, such as only sampling from the US or only from people who fill out online forms. However care must be taken to avoid or control for more directly relevant biases such as whether the twins are on good terms vs. estranged, whether they've taken a game theory class, whether they have the same profession, etc. The sample must contain at least 10 pairs. (A group of identical triplets can only count as a single pair, picking two of them at random.)

They're subjected to a prisoner's dilemma. (It must be clearly explained to them how the game works, so they understand the various incentives.) The stakes must be high enough that they actually care about the result; maybe a few hundred dollars. It's ok if they're able to find out about what their twin did afterwards. This market resolves to the fraction of twins (not pairs) who cooperate. If they're subjected to an iterated prisoner's dilemma, only the final game counts, and only if they knew it was the final game.