In many cases, the analysis that game theory provides does not seem to validate experimental observations. The best example of this can be seen in network exchange theory - where people simply refuse to accept what they believe are unfair offers. One of the classical explanations for this phenomenon arises from the Ultimatum Game, where it seems that people simply won’t accept some offers even if they get some money.

A simple experiment conducted by researchers at Princeton University (discussion) shows that we define fairness based on emotion - we balance anger with rationality to decide whether we accept an offer. The problem is that we cannot actually quantify our emotions on paper, and furthermore these emotions differ for each person. For example if we consider Alan Alda (person in the video), and say a childhood character from Slumdog Millionaire playing the same Ultimatum game (same opponents, same proposals), we can most certainly guess that Alan Alda would accept far less offers than a poor slum-dweller. This means that in essence, we accept offers as long as it does not make us terribly worse off than our opponent, weighed by social status. This explains the situation where if we could see that our opponent in the Ultimatum Game was a person wearing jewels and a fur coat (and was possibly some movie star), we might actually be more inclined to take those unfair offers - because we expect to take the amount corresponding to our social status.

This gave me an idea for a little variation of the Ultimatum game. Instead of playing against the other person in splitting the money, we can consider a game where the other person gives you the money, say as a token of goodwill. If you take the money only you get the money, but otherwise the money is just thrown out. If we consider a perfect player, he would obviously take all the offers that he gets. But if the reasoning above holds for humans, then we can consider what a normal person would do when given an sequence of very high offers by people of the same social status, followed by a low offer form a person of lower social status, he may actually reject that offer.

If we assume the this idea holds, we can actually start quantifying emotional impact based on some measure of social standing. What is interesting here though, is that while all the experiments are really ways of looking at the “basics” of social interactions, it seems our brain will not let that happen - we are programmed to work with social biases. In essence, attempting to strip the social aspect from interactions is almost impossible directly. This means that all experiments conducted with human subjects need to be able to isolate the “opponent”’s human aspect as much as possible. A good way to do this is to only be able to interact with your opponent with several options - disallowing the injection of any personality into.

While this may not mean a lot in terms of actual theory, consider what actually happens. An interesting example discussed previously here [1] and elsewhere [2], is the movie The Dark Knight. The passengers of the civilian ship all voted on detonating the prisoner ship, yet none of them were actually willing to do it. [2] shows this is a clear contradiction of theory, and explains it by introducing a moral cost to detonation. But the source of this moral cost is the knowledge of their opponent. Since the civilians KNOW that there are living people on the opposing ship, this causes them to alter their payoffs.

But in the end, there are always too many factors in play to understand the true nature of humans. While we may continue to dumb down experiments further into testing “basic” human decision systems, other factors quickly come into play. However we must remember that Game Theory itself provides the tools for analysis of almost any interaction - we just need to find out how to apply it.