How AI can play a catalytic role in managing criminal justice system for more equitable outcomes

AI has seen its applications across police departments and courtrooms in many countries. It has helped the administrative agencies to understand and decide the deployment of police officers at a particular location and also helped the judges in advanced countries in sentencing and granting bail.

As we witness the transition to 4th Industrial Revolution (4IR), while also slowly well equipping ourselves in the post COVID era, Artificial Intelligence (AI) can be seen to emerge as a plausible tool to enhance the effectiveness of law enforcement agencies across jurisdictions.

However, the question that concerns the administrative or government agencies is that: Whether AI can be used as a tool to ensure better criminal justice while policing? With the evolution of more sophisticated nature of crimes, is the conventional correctional system adequately prepared to deal with the crime and criminals? Have we been able to incorporate tech to tread the balance between retribution, rehabilitation and reintegration for a holistic sense of justice? (Swati Sudhakaran, AI in policing to reform the criminal justice system, ThePrint, March 2020) Are we going at the same pace with the rate of increase in cyber crimes? This piece is an attempt to weigh the advantages of using AI in law enforcement mechanisms and possible reforms in ensuring more equal outcomes in criminal justice system.

Indian experience of AI and its application in policing

AI has seen its applications across police departments and courtrooms in many countries. It has helped the administrative agencies to understand and decide the deployment of police officers at a particular location and also helped the judges in advanced countries in sentencing and granting bail. Recently, on the occasion of 70th Constitution Day, the Chief Justice of India Bobde, in his Constitution Day speech remarked that ‘AI can improve the judicial system’s efficiency through sophisticated and contextual automation of existing non-judicial tasks and functions’. (CJI 70th Constitution Day Speech, November 2019) This shall further help reduce pendency and expedite judicial adjudication.

In other phases of criminal justice administration in India, for example policing is replete with examples of AI applications in crime detection and resolution. One such example is of the start up called Staqu (Gurugram based) that has launched in November 2019 a video analytics platform- JARVIS or Joint AI Research for Video Instances and Streams, in Uttar Pradesh. ( The purpose of this software is to generate useful information out of the long CCTV video footage with short and crisp real-time alerts using AI and computer vision, thus significantly reducing the time it takes to come up with actionable data. Staqu is currently providing its services to eight states and union territories including Punjab, Haryana, Rajasthan, Bihar and Telangana. Similar application has been in case of Punjab Police (2018) where it has been actively using the Police Artificial Intelligence System (PAIS), also developed by Staqu. The feature of this product enables options like face search, text search and provides a database of more than 1 lakh records of criminals housed in jails across the state of Punjab. Product with identical features, called Trinetra has been also helping the UP Police.

Turning a leaf from the past, take for example- the Satyam scam (2009) when AI and machine was not at work so the police had to take a lot of effort to unveil the crimes committed by the CEO and CFO who had taken the advantage of dark hours to make false entries in accounting and showing dubious claims of profit. Whereas a crime of similar nature that was recently committed by a CA along with a tech expert of a power sector PSU to manipulate the accounts during the late night hours in which only a computer to detect the difference between the Zero and the alphabet ‘O’, was brought into light due the installation of machine learning that sent alerts for suspicious activities. This was further corroborated with the help of CCTV footage that caught these two employees. Therefore, AI can help in detection of advance crimes today.

International experiences of application of AI in policing

Some of the successful examples of computational methods of ‘predictive crime mapping’ has been implemented in United States, almost 12 years ago. US has used CompStat (Computer Statistics, or Comparative Statistics), geospatial modeling for predicting future crime concentrations and has developed into a paradigm of managerial policing while employing Geographic Information Systems (GIS) to map crime. (Ales Zavrsnik, ‘Criminal justice, artificial intelligence and human rights, February 2020) It relies more on humans to recognize the patterns. The disadvantage in such cases, is that it focuses on the ‘surface’ and the not the ‘cause’ of the crime. However, it does help in crime reduction, improved quality of life and personnel and resource management.

Moving from ‘computational mapping’ to ‘predictive policing’ has helped the police in processing the depth of information and pull out operationally relevant knowledge from the data collected. AI has been instrumental in helping the police penetrate into the preparatory (ex-ante) phase of crime, yet to be committed and also act ex-post-facto, after the crimes have been committed. For example, the Interpol, in Europe manages the International Child Sexual Exploitation Image Database (ICSE DB) to fight child sexual abuse (Interpol website). It has helped identify the criminals and victims with the aid of furniture and other mundane items in the background of abusive images. Interpol on April 9, 2020 has released international guidelines in order to enhance safety and effectiveness of law enforcement in the context of COVID-19 pandemic. In the wake of the pandemic, cyber criminals have evolved tactics, techniques and procedures (TTP) to exploit vulnerabilities. In such cases, chatbots acting as real people can be used to identify the offenders. This can be done without human intervention to not only identify persistent perpetrators but also deter the first-time offenders.

Using AI for effective prison management

The Indian prison management is complex given the crowded nature and the current infrastructure in place. AI can, as such, help in reducing the stressful environment by effectively managing the police personnel and prison inmates. It can be helpful in tracking both- abusive behavior by the police and the prisoners to better map and chart out a path to deal with the miscreants. It can effectively manage drug trafficking issues in jail and enhance the role of prison as an appropriate correctional facility mechanism. AI can also help in deciding cell allocation based on age, family background, and nature of crime committed and criminal history of the accused or convicts.

AI and its challenges to the criminal justice system: Way forward

AI based products have revolutionized the concept of modern security with state-of-the-art technology and intelligent monitoring of objects, crowd, perimeters and vehicles. With the evolution of AI systems (big data analytics, machine learning), the assessment of risks and operation of criminal justice system has become increasingly technologically sophisticated and may also leave important consequences to the criminal justice system if not used wisely. It is important to understand that AI works on human biases and stereotypes based on the data fed into these algorithms by humans. There is a high probability of reinforcement of historical disparities if due caution is not exercised at the very first stage. Thus, there is a need to find a way out to make it more neutral in order to make dispassionate decisions about policing and punishment (Katie Brigham, March 2019,

Unlike advanced countries like US, India has yet not witnessed a wide usage of AI to be able to delve into the deeper challenges it poses. In the US, the judges to grant bail and also while sentencing have relied on AI. However, the ‘risk of recidivism’ i.e. the likelihood of reoffending, remains subjective and is something that cannot be captured with accuracy with the help of AI. A possible solution could be to increase the accuracy of AI to better capture human biases to the extent that it becomes minimal or negligent. This can be done, if the software making companies does a thorough research into the bias in AI and the impacts of using AI in criminal justice administration. It is also important to include diverse stakeholders in creation of the software so that multiple perspectives are mapped to the algorithm.

AI is here to stay and this challenge is not only for the criminal justice system but also for the tech industry as a whole. While developing software, it is essential to take into account the biases felt by individuals who have been arrested or imprisoned such that the efficacy of the AI leads to more accuracy and transparency, both for the police and the courts. There will be trade-offs but the potential of AI in reducing crime cannot be overlooked. A pragmatic approach would be to appreciate the shortcomings, work on the loopholes for the betterment of criminal justice systems as AI evolves and helps in shaping the society.

Blog by- Dr. Parvez Hayat

Former DGP/ Advisor Centre for Climate Change and Disaster Management JMI -A-Central University