Imagine that you have been incarcerated for committing a crime. You get to choose one of the prisoners as your roommate, but your choice is limited to a thief and a murderer. Who would you trust more?
Most people would consider the thief’s crimes to be lighter and choose him without a doubt. They would rather live with someone who stole some jewellery than some beast that killed another human being.
But if you think about it in depth, you may come to the opposite conclusion. Stealing is often a meticulous and calculated crime that involves logical planning, but murder is more often a crime of passion that is spontaneous (of course the person may be a psychopath). Ergo, statistically speaking thieves tend to be more predisposed to a criminal nature than murderers. A thief will easily repeat their crime, whereas those who have murdered often do not repeat it. Furthermore, thieves tend to target people they do not know, where as murderers often kill someone that they know. Statistically, you are more likely to die in the hands of someone you know well than a stranger.
Therefore, it just might be that sharing a room with a murderer is preferable to a thief (maybe the murderer will not steal all your secret food). This would definitely be the case if the “murderer” was actually framed like Andy Dufresne in The Shawshank Redemption. Of course, the wiser choice still is to not commit the crime in the first place so you do not end up in prison.
The prisoner’s dilemma is a famous example of how game theory functions. It predicts the behaviour of two people when forced to cooperate. The story goes as follows:
Two accomplices in crime are arrested by the police. They are interrogated in separate rooms. As the police have insufficient information, they offer a deal to each prisoner to confess that the two committed a crime (or deny). The deal is:
If you confess and your partner denies taking part in the crime, you go free and your partner will serve ten years (and vice versa).
If you both confess you will go to prison for four years each.
If you both deny taking part in the crime, you both go to prison for two years.
Assuming the prisoners act rationally (i.e. for their best interest and minimising their jail time), the prisoner will obviously choose the “confess” option as this is hypothetically the best choice (minimum time = 0 years, compared to only 2 years minimum for denying). However, because both prisoners are thinking this, the result is almost always that both confess and end up with four years each. Therefore, because human beings are unable to trust another human being enough, people always end up acting irrationally (benefit not maximised). If the two had been trusting (assuming the other would deny too) and cooperated, both would have served half the time. But people always assume (correctly) that the other person will betray them for their selfish gain and this win-win result is unattainable.
But what if the other prisoner was yourself? Let us assume that the prisoner’s dilemma game was played by you and an exact copy of you. A copy that thinks like you, acts like you and identical to you in every single way. Can you trust yourself? Do you trust yourself enough to deny the crime, when it is entirely possible that he or she rats you out to walk free while you suffer for 10 years? How do you know that he loves you more than himself?