Yes, winter term...'tis the season to moan and groan about comps* projects. Jaime Olson '07, as a double major in cognitive science and computer science, has two comps to contend with this term!
His cognitive science project involves conducting a study in the field of game theory. You may remember that scene in A Beautiful Mind where John Nash is strategizing with his friends in a bar on whether to go for the blonde or the brunettes. That's game theory. Players are engaged in some sort of competition, and each is trying to get the best outcome for himself or herself.
Jamie's game theory project uses a modified version of the Prisoner's Dilemma as the game. The story behind the Prisoner's dilemma is this: Two men have been detained on suspicion of burglary, and are being held in separate cells in the police station. A police officer comes to one of them and tells him that they have insufficient evidence for a burglary charge. But if he will confess to the burglary, he will get off scot-free for cooperating with the police and his partner in crime will get five years in prison. But if his partner confesses, too, they will both spend a year in prison. If neither of them confesses, they will both get three months for disorderly conduct. Does the burglar confess? It depends on how well he trusts his partner.
In Jaime Olson's version, instead of taking place in a police station, the game is played at a computer terminal. The students who volunteered to participate are given the choice either to "cooperate" or to "defect." To defect is equivalent to ratting on your fellow burglar. Meanwhile, at another computer terminal, your opponent makes the same choice. Points are awarded as follows:
- If both students choose to cooperate, both receive 3 points.
- If one student chooses to cooperate and the other chooses to defect, the defector receives 5 points and the cooperator gets 0 points.
- If both students choose to defect, both receive 1 point.
Here's the twist: after several rounds are played, each player's game history is displayed on the bottom of the screen. Not only do players have the opportunity to strategise their next game move based on their opponent's history, they must also weight the fact that their own game history is publicly available. If you defect too often, your opponents will no longer trust you. The object of Jamie's study is to see what happens to "cooperation if your future moves matter."
Not content with simply studying human subjects, Jamie is also working on computer software that simulates virtual Prisoner's Dilemma players. "You create a population of artificial agents that play like the people here," he says, "then use a genetic algorithm to kill off the poor agents." In plain English, Jamie's program creates a large number of game "players," that are little bits of computer code. The game players that lose too much are eliminated, while the ones that win are used as templates to generate new game players.
Jamie's computer science project is on a totally unrelated subject. "It's a lot easier," he says. Computer science students work in teams on a certain project assigned by the advisor. Jamie's team's task is to build a natural-language grocery picker program. This program would lend a user assistance in deciding what groceries to buy for the upcoming week and would be able to understand English rather than computer code. If you have ever been frustrated trying to get Microsoft Word to understand what you want when formatting a document, you will know that this is no easy task.
Still, Jamie does not seem in the least fazed by his double comps project. If anything, he may prefer the cognitive science project because of the freedom students get in selecting an area of study. "You get to pick a topic you're interested in and run with it."
*For those of you who aren't Carleton seniors, comps is the huge paper or project that you do in the area of your major as a capstone to your Carleton experience.
