Prisoner's Dilemma
Survey Report


The Prisoner's Dilemma is a standard example of a game analyzed in game theory that shows why two completely rational individuals might not cooperate, even if it appears that it is in their best interests to do so. It was originally framed by Merrill Flood and Melvin Dresher while working at RAND in 1950. [Source: Wikipedia]

This survey was a modified simulation of the same. It is a part of my Philosophy Project at college. This report shows the results and observations for the same.

This survey received around 150 responses in the first week of being live. A big thank you to all those who contributed! :)

The Original Case

In this case, the penalties were set to resemble the original Prisoner's Dilemma (which however assumes the other person to be intelligent only).

Though, we do get to see rather mixed opinions with only a little cooperation/betrayal margin, opposed to the original paradox, according to which number of betrayals would be a lot more than number of cooperations.

The additional thing here is the assumption of the other person being unintelligent. We can observe a deviation in the distribution in this case. People tend to Betray when they assume that the other person is not intelligent.

Putting it differently, we may say that people do tend to cooperate if they assume that the other person is intelligent, probably expecting for a similar action on the other end.

When Penalties were Reduced

In this case, the penalties involving at least one of the two people cooperating was reduced by a great extent.

It is very evident from the responses that people prefer cooperating in this case. This shows that people are willing to cooperate as the risk factor reduces. It is also worth noting the order in which the cases are put up. That is, people tend to cooperate in this case after they faced the original situation in which the penalties were higher.

Also, it can be observed that the number of betrayals increases when the other person is assumed to be unintelligent. More on this in the last section.

When Cooperation was Favoured

In this case, mutual cooperation was favoured, by releasing all charges when both cooperate. Also, in case of a conflict, where one betrays and the other cooperates, even the one who betrays gets some punishment (in contrast to the original case where the person who betrays is set free in such a situation).

A very good thing to note is that this simply works as expected :) There are very less betrayals compared to cooperations.

Again, the ratio of number of betrayals to cooperations is higher in the case where the other person is assumed to be unintelligent.

When Some Penalties were Unknown

In this case, the punishment for common responses (either both cooperate or both betray) were kept ambiguous. This case was intentionally put up in the end, in an attempt to build up a mindset that generally mutual cooperation leads to lesser punishment. In case of a conflict (one cooperates and other betrays), the one who cooperates suffers larger punishment.

The motive here was to see whether people cooperate based on an instinct built up from the previous cases or they betray just seeing the lesser punishment (and assuming the other would cooperate).

One interesting thing to note is that many people avoided this case since there were considerably lesser (around 20 less) responses to this case when compared to other cases.

For the responses received, we can observe that people tend to cooperate when the other person is assumed to be intelligent. However, in the other situation, the difference between number of cooperations and betrayals decreased.

That is, people tend to go with the instinct instead of the apparent benefit (which however involves some amount of risk in case the other betrays).

Intelligent or Not?

This is a comparison for total number cooperations or betrayals when the other person is assumed to be intelligent versus when the other person is assumed to be unintelligent, summed up over all cases.

It can be easily observed that number of betrayals get increased by a lot when the other person is assumed to be unintelligent.

This is an interesting point. Let's say we fix that person A betrays. We have two cases. One, A thinks that the other unintelligent person (B) would cooperate, since B is not intelligent, and hence A would betray for getting released. Two, A thinks that B would betray. But in this case, A should prefer to cooperate in order to get lesser penalty (unless A doesn't want a tit-for-tat). So, A won't betray.

Hence, people tend to betray more under the assumption that the other person is not intelligent, thinking that the other person would cooperate.


As I mentioned before, Prisoner's Dilemma is a well-known paradox and has many use-cases. This modified simulation of the same had two main objectives:

One, to see if the present responses are in line with prior observations.

Two, to observe and analyse the affect of modification in parameters like rationality and reward (rather punishment in this case).

The observations in prior sections catered to these objectives.