New research into network science suggests that with sufficient data, it may be possible to predict which politicians are likely to become involved in corruption before the crime actually happens.
Researchers from Brazil’s Maringà, Sao Paolo and Ponta Grossa universities and Slovenia’s Maribor university compiled known and verifiable cases of political corruption that took place in Brazil over the 27-year period between 1987 and 2014 – a dataset which, thanks to Brazil’s history of political corruption, contains more than 400 individuals who can be shown to have been involved in 65 “important and well-documented” scandals.
By linking these individuals where they had been involved in the same scandal, and analysing the structure of this “corruption network”, the researchers deployed algorithms to see if they could say where the “missing links” would turn up from one year to the next. “Some of these algorithms”, they found, “have a significant predictive power”.
Despite the relatively small dataset, the best algorithms applied were able to predict with more than 25% accuracy which links were going to appear in the corruption network, compared to an accuracy for a random approach of 0.2%. The researchers also pointed out that this is likely to be an underestimation of the predicting power of the network as many corruption cases go uncovered.
The science of making predictions from network connections hit the headlines in 2013 when data scientists from Facebook and Cornell University analysed 1.3 million randomly selected Facebook users and the 8.6 billion links between them. Algorithms applied to this network could say with 60% certainty who was married to whom, and could also predict which relationships were likely to fail. The story, as widely reported, was that Facebook knew whether a user’s marriage was going to end in divorce before they did.
Political corruption siphons off more than 5% of global GDP, with over a trillion dollars in bribes paid each year worldwide. While predictive “detection” of any crime raises serious ethical questions, this research could offer useful insight into the way in which political corruption takes place, and how systems might be designed to mitigate or prevent it.
The full paper, entitled The dynamical structure of political corruption networks, can be found here.