Game Theory vs. Crime
Link: https://engineering.vanderbilt.edu/news/2014/game-theory-can-help-predict-crime-before-it-occurs/
Crime will soon be a distant memory. Referencing the movie Minority Report, this news article on the team’s research states their goal is to predict crime and be prepared to stop it, just like in the film. This seemingly far-fetched achievement is already being put into practice. At Vanderbilt School of Engineering, a team headed by Associate Professor Eugene Vorobeychik is stopping crime before it happens. They’ve used game theory based on crime data, weather, and the experiences of the local police to optimize policing. Through analysis of big data, more officers are sent to areas likely to have crime, which efficiently allocates patrols and resources. These predictions aren’t perfect though. While Minority Report had psychics, who could see the future to reliably stop murders, our world’s best bet is to produce models based on probability. This is can be seen as a real-life application of Cornell’s Networks course.
In the content of Cornell’s Networks, game theory is taught with an emphasis on Nash Equilibrium. We’ve discussed pure strategy Nash Equilibriums, which rely on each player in a “game” having a best response to each player’s choices. Sometimes, there is no best response, so some probability will be mathematically factored in based on external factors and payoffs. Those are mixed Nash Equilibriums. That’s why the finalized, chosen model of Vanderbilt’s research could be interpreted as a Mixed Nash Equilibrium since crime prediction wouldn’t give a pure strategy equilibrium. Either the police win or crime does. Payoff values based on available information will influence the values of p and q in mixed Nash Equilibrium calculations. The model that prevents the most crime will be implemented in the city, which shows how flexible game theory can be for real-life situations. From traffic to stopping crime, Networks shows up in real life.
Here’s an example of how a Mixed Nash Equilibrium could be generated for allocating patrols to locations, where each “player” would want the highest payoff. In this case the numbers for police would be crimes stopped and for crime would be crimes succeeded; Site B, Site B would be chosen most often.
