Remember the science fiction movie called Minority Report that starred Tom Cruise where future crimes can be predicted by a system thereby helping the police to catch the criminals much quicker? This far-fetched concept has now become a reality. The role of artificial intelligence in our day to day lives is unbelievably vast. From getting recommendations on what winter coats to buy this chilly season while scrolling through your Instagram to seeing advertisements of food delivery apps that you use frequently, before watching a Youtube video, artificial intelligence has seeped into every aspect of our digital lives. And now, it is going to permeate itself into the justice field as well.
Predictive policing borne from big data analytics is a kind of a law enforcement technique where algorithms are used to convict criminals and predict prison sentences. It uses data and statistical analysis to aid in the identification of criminal activities and its major objective is to help reduce the crime rates by providing the police forces with details of high risk areas. Predictive policing is still in its nascent stage and it has to be understood that machine predictions can never be hundred percent accurate. For this very reason, the usage of predictive policing in justice systems raises several ethical concerns.
For example, in the United States, there have been several cases wherein algorithms used by different companies as a tool for predictive policing have predicted that a Black individual has a higher probability of committing a crime than a White individual. Such biases in the system might be dangerously unfair to these individuals because they will be given a longer jail time if they get convicted as criminals. So, where do such biases come from? The answer is rather simple. These biases stem from the data that is being fed into the system. Historically and statistically speaking, in the United States, the case has always been that the number of Black individuals committing crimes is higher than the number of White individuals. This biased data feeds biased algorithms and hence, the results are slightly skewed. The system is going to pick on the fact that in the data that it contains, majority of the criminals are Black individuals or immigrants. Similarly, if the system is incorporated into other fields, it will start noticing a trend of how in most places the role of a receptionist is reserved for a female and that a doctor is predominantly a male. The information that is being fed into the system, no matter how true it is, will result in unavoidable biases that make the predictive policing system very dangerous.
Another problem with predictive policing is that it usually identifies areas stricken with poverty, low education and ethnic communities as high risk areas. This is not necessarily true and it can be very misleading. Several surveys that where conducted show that there need not be a correlation between poverty and crime rates in many places. Just because some places have that correlation, the system cannot apply it to every other place. However, the system inevitably projects the results of one neighbourhood to predict the results of another, hence it results in several biases and skewed judgements.
In October, the International Conference of Data Protection and Privacy Commissioners released the Declaration on Ethics and Protection in Artificial Intelligence which is a set of principles to govern AI. It hopes to reduce unlawful biases and discriminations and that AI should be evaluated on a broader set of legal and ethical criteria. Predictive policing is a great technique in identifying criminal activities and the like. In the future, it might serve as a great tool to smoothen out several tedious judicial processes. However, it is not with its disadvantages. Research in finding new algorithmic techniques to identify misrepresentation, make race neutral predictions and to make effective counter measures to detect manipulated data in the media can prove to be useful in mitigating such biases and in making machine learning predictions more accurate.
Picture Credits : online.campbellsville.edu