Once upon a time, criminals had no borders. They would pick places at will and Hindi cinema would glorify them with lines like, “Don ki talash toh gyarah mulkon ki police kar rahi hai, lekin Don ko pakadna mushkil hi nahin, namumkin hai’’. The world was also a more interesting place before facial recognition technology was deployed. Now, across airports, bus-stations and traffic signals, cameras are ubiquitous.
Once everything is connected, this technology can identify people using any camera in the world. The technology is far more advanced than the people it caters to, making the possibility of a surveillance state not too far off in the future. Thus, it is the right time for the US Democratic lawmakers who introduced a bill that would ban the use of facial recognition technology by federal law enforcement agencies. But, do we need something similar in India?
According to reports, the US bill would make it illegal for any federal agency or official to “acquire, possess, access, or use” biometric surveillance technology in the US. Law enforcement agencies leverage and utilise facial recognition technology to identify victims and suspects by matching their photos and videos with pre-existing databases, such as driving license records and other markers of identity. But civil liberties groups have been raising concerns about facial recognition being prone to privacy erosion and reinforcing racial biases.
Lawmakers argue that technologies like facial recognition violate human rights and privacy and the discourse needs some intervention before we suffer a data breach and an unprecedented crisis.
DEMOCRATISING AI: A DOUBLE-EDGED SWORD?
The proposed law has arrived at a time when the use of facial recognition technology by law enforcement agencies came under major fire amid the protests against the killing of George Floyd in late May this year. Studies have repeatedly shown that facial recognition is less accurate for people of colour, specifically the Black community. The New York Times reported a similar incident wherein an innocent black man in Michigan was arrested due to misidentification by a facial recognition software.
The exact reason for the misidentification of darker-skinned people by facial recognition software is still unknown. One of the main factors is that the data being used to train algorithms is said to have a disproportionate number of white males as opposed to people of colour and women. Another reason for the misidentification by law enforcement agencies is that facial recognition software uses mugshots from criminal databases to find an exact match and the databases have a disproportionate number of people of colour, leading to false matches. The US criminal justice system functions with the aid of a computer programme for sentencing and granting bail. It has been observed by data scientists that the system is disproportionately unfair towards people of color. As reported by journalists Harry Armstrong and Jared Robert Keller in The Guardian, “In the US, criminal ‘risk assessments,’ based on predictive analytics, have already been shown to be biased against black individuals because the data used to build the system is inherently biased.”
AI is inherently unbiased and what it learns is purely based on the data we provide. So, we must be careful while preparing the dataset before training AI. Even in case of seemingly perfect data, there might be a bias, based on the features we ask the AI models to learn or pick from the dataset. So, it is critical to constantly analyse the results in order to improve the dataset. It is also important to put in place periodic updates and additions to the dataset as well as re-train the AI models.
From the Indian perspective, we should learn from the experiences of others and ensure that we do not repeat the same mistakes. If used correctly, AI can help improve people’s lives.
India’s judicial system is a textbook case of ‘justice delayed is justice denied’. It has a complex mix of problems. We have one of the most understaffed police forces in the world. Each state has its own police force that faces political pressures on a day-to-day basis and has to work with archaic laws and an infinitely long cycle of judicial hearings.
Deploying facial recognition technology will help the forces in critical areas. The central government has approved the implementation of Automated Facial Recognition System (AFRS). It will help police forces to search for criminals and missing and dead people from a nationwide database.
In our country, there is no precedent of law enforcement agencies being biased towards darker-skinned people. However, there are different parameters for discrimination in our police forces, based on caste, economic status and, sometimes, religion. Therefore, the bias of a facial recognition system for darker-skinned people is not relevant to India. But, the threat of a privacy breach and this powerful tool being misused still remains, which is why we need very strong laws to protect our citizens.
REVAMPING POLICY FRAMEWORKS
Independent data regulators should build a progressive data regime. The right steps to protect the data and its misuse need to be taken. At present, there are seven international agreements and standards centred on data privacy, out of which, five require the establishment of an independent supervisory authority. The Organisation for Economic Co-operation and Development (OECD) Principles do not stress on the need for an independent supervisory authority. However, with the Europe model, both the GDPR (previously Directive 1995) and the Convention 108 of the Council of Europe i.e. 90% of countries with data protection laws have opted for this model.
India introduced the Personal Data Protection Bill in December 2019. It is a thorough document covering most of our concerns. Currently, the bill is still under review by the committee and stakeholders. If it gets slated for early next year, we can have an ombudsman of data in place, at the earliest.
The bill has provisions to exempt government agencies from the bill in certain circumstances involving national security, in accordance with the procedures established by the law. These are a few grey areas we must be cautious about and make sure that the laws are put in place keeping in mind both the privacy of people and the security of our nation.
THE WAY FORWARD
In both the Indian and global contexts, we need to ensure that facial recognition technology is used ethically and does not violate the privacy of end-users. General consensus among the public is mixed. For instance, the Bangalore airport is implementing face recognition based on boarding passes and not facing any repercussions since its use is restricted to the airport with a choice to opt-in or out of it. The government promotes CCTV camera usage in public places for safety purposes and the reaction of the public is generally positive towards it too. In fact, in all such areas, it is repeatedly announced that people are under CCTV surveillance.
Before introducing a new technology that will redefine how our law enforcement agencies function, due diligence should be done. Transparency in acquiring data and its uses will help the general public understand how facial recognition is helping to create a safer society. There should be open debates about the threats along with the benefits such a system brings to our police force. Rules and regulations around the use of technology should go through proper channels.
Biometric data has simplified life for several people across the country. To minimise the misuse of facial recognition, it should be under the control of a single government body which should be made answerable. It would also reduce the blame game in case of a breach and make the process of finding loopholes easier. The state agencies should work with central agencies to use the resources available, where predefined quotas will help remove discrepancies between the usage among different state agencies.
Meanwhile, the starting point could be a system like the face recognition-based boarding pass at BIAL, where the public can decide if they want to opt into the system or not, just like the option to lock or unlock Aadhaar details and biometrics.
Facial recognition is a very powerful tool that can reform law enforcing processes in India. It will help locate criminals in a short duration, making the society a safer place. Eventually, cameras will identify all criminals and no ‘don’ will ever be able to make his boastful statements ever again.
The writer is CEO and co-founder of Integration Wizards Solutions.