A research team led by scientists from the Center for Complexity Science in Vienna, a research institution that aims to collect and coordinate research on complex systems, has found that big data can help predict crimes statistically.
According to the study published by this team in the journal Scientific Reports, the team studied data on 1.2 million crimes that occurred over a period of 6 years in a small country in Central Europe. They classified all crimes into 21 categories, including corruption, terrorism, theft, sexual crimes, and others.
Behavior of criminals
By studying these crimes, the team concluded that criminal offenders who specialize in certain types of crimes are usually female and older, compared to individuals involved in a broader and more diverse range of crimes.
The study indicates that these specialists also tend to operate within a more limited geographical area and therefore rely on local knowledge, their criminal connections are usually within the same city, and they operate in more cohesive local networks, which increases the likelihood of repeat partnerships and crimes.
On the other hand, it appears that those who commit a greater number of crimes and do not specialize in specific crimes have larger and more open networks.
The study also found that transition between certain types of crimes occurs significantly more than others. For example, perpetrators of crimes against children tend to shift to work in prostitution, crimes of cruelty to animals, or computer-related crimes (such as data theft, blackmail, and electronic fraud, for example).
This style of studying criminal behavior could help law enforcement agencies better anticipate criminal developments, he said statement Official press release from the Center for Complexity Sciences in Vienna.
This research pattern has already been observed before. For example, in cities such as London, the police build experimental maps that are updated on a daily basis, including the types of crimes occurring on the streets, the network of active suspects, the institutions and shops most likely to be exposed to robbery and looting, etc., and this data is placed in algorithmic engines to predict Where will the next crime take place?
In the United States of America, New York police departments apply experimental versions of prediction-like methods, where algorithms capable of self-learning through artificial intelligence are fed with data on citizens in a city, their criminal records, and the dates, locations, and nature of every crime that occurred in the area. Here, the algorithms give results about the categories most likely to be deposited in… Prisons and places closest to beatings.
This new approach is called “conditional prediction,” but it faces a major ethical problem, as there are specific societal groups that are highly exposed to pressure because of their gender, race, or mental state.
Take blacks in the United States, for example, which is the most studied example. It is known that the number of black and white youth who use cannabis is almost the same, and yet the number of cannabis users who are put in prison tends to be black, with a difference of 3. Up to 4 times.
There are major political and social reasons for this, but the algorithms do not pay attention to them. Once they are fed data, the algorithm will immediately learn that there is a tendency for blacks to be more likely to be imprisoned for minor crimes, and the same applies to other groups such as immigrants.