Author : Yashasvi Maharshi from Rajiv Gandhi National University of Law
For centuries, constitutional democracies feared the concentration of power in kings, dictators, and institutions. The twenty-first century introduces a different danger: power hidden inside algorithms.
What happens when governance no longer speaks through judges, bureaucrats, or legislators but through opaque algorithms that nobody can fully explain?
Artificial Intelligence is increasingly being employed by governments across the world to make decisions that were earlier left to humans. In addition to police work, algorithmic decisions are being made in regard to welfare provision, tax collection, border control, recruitment, and management of courts. Constitutionalizing algorithmic governance is the current reality in India due to the rapid growth of digital architecture for governance in that country.
However, the employment of such technologies poses a serious constitutional problem in relation to the issue of accountability in democratic governance when human institutions give way to automatization. The aim of the Indian constitution is control over human power. Notions of accountability, transparency, reasonableness, equality, and procedural justice are all constitutive elements of constitutionalism. Nevertheless, constitutionalism faces novel challenges when AI operates in mysterious ways that even its inventors cannot explain.
Therefore, the issue at hand cannot be resolved solely through technological means. Instead, it is fundamental to constitutionality. An era where citizens are governed by opaque computer machinery rather than visible and accountable mechanisms may be ushered in by the phenomena of algorithmic governance. Efficiency may rise under this plan, but constitutionality may progressively deteriorate.
Understanding Algorithmic Governance
Algorithmic governance involves the application of data-centric computational models for aiding, influencing, or automating governmental decision-making processes. Contrary to conventional administrative systems, the algorithmic process employs prediction models, machine learning, and huge data sets to create outputs.
These systems are increasingly used in areas such as predictive policing, surveillance, welfare administration, digital identity verification, public service delivery, judicial analytics, and taxation systems thereby expanding the role of automated decision-making within the administrative state.
There is an increasing integration of technology in state administration within India as demonstrated by the use of digital governance through technologies like Aadhaar, facial recognition, and AI-led administrative practices. Although facial recognition technology has been employed in protests and events, the police force has experimented with several technologies for predictive policing. Moreover, the Indian judicial system has tested technologies for case management and language translation enabled by artificial intelligence.
Supporters of algorithmic governance suggest that there will be increased efficiency in system operations, reduced corruption, and easier policy implementation due to AI. With regards to implementing automation through algorithmic governance, there is reason to see that idea implemented as a result of the large population numbers and absence of bureaucracy. It is much faster for data to be analyzed by algorithms compared to bureaucrats.
However, efficiency is not the only aspect of governance. Constitutional democracies are in place not only to expedite processes but also to guarantee the equitable use of power. For this reason, there are constitutional issues with algorithmic government.
Black-Box Governance and the Crisis of Transparency
Opacity is one of the main problems with AI-led governance. Many AI systems function as “black boxes,” which means that regular people are unable to comprehend the underlying mechanisms. This kind of problem also arose during the issue of Aadhaar-based welfare exclusion, where the beneficiaries had to be barred from accessing essential services owing to problems related to technology. In most of these cases, it was hard for those who faced exclusion to figure out the underlying logic behind their exclusion. There is a situation known as “procedural invisibility” in which individuals are unaware that algorithms were used to make choices about them.
Let’s examine an instance where automated data analysis prevented a welfare client from receiving services. The individual won’t comprehend:
- why the decision was made,
- what data was used,
- or how the conclusion was reached,
then the ability to challenge state action becomes severely weakened.
In fact, it directly jeopardizes the principles outlined in Article 14 of the Indian Constitution. The Indian constitution has long emphasized that the government should not act arbitrarily. Cases like Maneka Gandhi v. Union of India, in which the Supreme Court modified Article 21 of the Indian Constitution to allow for procedural reasonableness, demonstrate this. According to constitutional law, every action taken by the government must be logical and understandable.
However, algorithms present a risk since they substitute probabilistic computing for the process of reasonableness in decision making. “The data says so” has replaced “the law says so” in decision-making. This indicates that the threat is not just technological but also constitutional.
It is difficult for someone to seek compensation for something they do not comprehend. Constitutional responsibility is predicated on the existence of power. However, an uncertainty of accountability and responsibility arises when choices are made using automated procedures using complex data sets.
Algorithmic Bias, Surveillance, and the Rise of the Predictive State
AI technology is assumed to be free from bias and objectivity in its nature. Nevertheless, what is problematic about the algorithmic technique is that it merely automatizes the biased history that precedes but does not eliminate it. As there is regional discrimination, gender discrimination, a hierarchical caste system, and a biased police force within our society’s history, it is quite likely that the algorithm will show its discriminatory nature.
There is a certain danger since artificial intelligence technology can spread across different sectors, such as surveillance, welfare administration, recruitment, policing, and many others. For instance, if police officers use predictive software, they will continue targeting specific neighbourhoods due to the previous biases connected to high crime rates in particular areas. By analogy, the algorithmic recruitment process will be biased against marginalized groups since it will follow the old hiring practices. Studies have demonstrated that the accuracy of facial recognition algorithms drops dramatically when it comes to recognizing women and minority groups. The problem of algorithmic bias acquires more relevance in India, considering that in recent years there has been some experimentation on the part of various state governments to deploy facial recognition technology while there are protests and large assemblies in the public sphere. Concerns about the misuse of facial recognition technology were heightened due to its use during the protests in Delhi in 2020.
Article 14 of the Indian Constitution, which prohibits state arbitrariness and guarantees substantive equality, is threatened by the sheer existence of such systems. Algorithmic prejudice, in contrast to historical discriminatory practices, is concealed by its complexity and rationalized by efficiency. The individuals may never be aware that an algorithm had an impact on the choice that altered their lives. Because constitutional violations can become statistical truths in such a scenario, it is possible to refer to it as procedural invisibility.
Simultaneously, new artificial intelligence-based technologies are altering the character of the citizen-state interaction through surveillance. The government has previously unheard-of opportunities to monitor people thanks to facial recognition technologies, biometric identification systems, predictive analytics, and profiling tools. It became evident that privacy is a component of freedom and dignity with the historic case of Justice K. S. Puttaswamy v. Union of India.
In fact, the creation of the “chilling effect” is the biggest danger modern technology poses. The premise here is that people tend to police their own behaviour in order to stop any kind of state intervention before it is truly necessary since they feel like they are being watched all the time. Thus, it becomes evident that surveillance is subtly weakening democratic participation through psychological manipulation rather than explicitly damaging it.
More significantly, AI transforms government from a reactive to a predictive system. This suggests that rather than just responding to citizen wrongdoing, the government makes an effort to forecast future misconduct using risk assessment techniques and specific probabilities. One could refer to the emergence of the “predictive state,” in which the suspicion of wrongdoing arises not from actual wrongdoing but rather from the expectation that it might occur in the future.
What this suggests is the potential for a constitutional democracy to transform into one where the government no longer governs its citizens through laws, but through algorithms.
The Computational Governance and India’s Legal Vacuum
AI, however, brings a whole new perspective, not only to the working of the government but to the very core of the concept of governance. Constitutionalism has always rested on the idea that the exercise of governmental power would be directed towards certain human beings who could be held responsible for their decisions and actions and would be open to transparency, accountability, and oversight. In other words, the entire concept of governance rested upon visibility, reasoned authority, and democracy.
The use of technology and AI will result in a transfer of power from governmental organizations to the use of computation systems based on data analysis and probabilistic calculation. As governments start using algorithms designed by engineers and computer scientists rather than judges and bureaucrats, the entire logic of governance would undergo drastic changes.
This is because, while AI technology focuses on efficiency, speed, optimization, and accuracy, constitutionalism stands for values such as accountability, reasonability, deliberations, equality, and respect for an individual. In other words, AI will not only facilitate governance but also transform the government itself.
Nevertheless, this risk is subtle and yet enormous. Democratic degeneration does not take place overnight through a shift to authoritarianism, but instead gradually, through successive administrations that could seem technologically progressive and administratively proficient but had actually destroyed democratic systems within. It would be possible then, for algorithmic governance to give rise to such a regime – one which would be technologically proficient and yet unconstitutionally confusing.
Despite all these consequences of the technology, currently India does not possess a comprehensive policy framework regarding artificial intelligence and automated decision making by the government. All regulations till date, be it the discussions held by NITI Aayog or other digital governance policy decisions, have all been innovation-driven and not human rights-focused. Moreover, while the Digital Personal Data Protection Act does tackle some aspects of informational privacy, it does not fully consider algorithmic accountability and violations by states through automated decision making.
Alternatively stated, currently India has no regulatory frameworks that seek to categorize algorithms and their deployment in government operations. In contrast to the EU Artificial Intelligence Act which seeks to categorize these algorithms, there are no measures for regulating the transparency and accountability of these algorithms. This lack of regulation is especially problematic because it is difficult to prove that harm done by algorithms exists; thus, violations of constitutional guarantees might go unnoticed until automation is embedded into the process of governing itself. With no constitutional protection in place, India stands at the brink of the age where the exercise of public authority is no longer institutional but computational, remaining unnoticed by the people.
Reimagining Constitutional Safeguards in the Algorithmic Age
Thus, there is a need in India to develop a constitutional framework that would help govern the power of algorithms and yet not limit innovation.
First, citizens should enjoy the right to be informed when algorithms are involved in governmental decision-making, especially where they touch upon their entitlements or rights. In such instances, individuals should have information on whether AI has been used in making such decisions and the process through which the decisions were made.
Secondly, algorithm impact assessments should be carried out before the use of AI in governmental processes. The assessments are intended to determine any potential harms related to issues of privacy, civil liberties, discrimination, and the protection of the Constitution.
Thirdly, high stake decisions by the government on areas such as access to social services, national surveillance, criminal justice, and individual freedoms should always be subject to some kind of human supervision. Constitutional governance cannot afford the use of automated processes in making decisions.
Fourthly, an appropriate framework for carrying out constitutional audits of AI is necessary. Currently, most audits are carried out with respect to the issue of security. However, constitutional audits of AI involve an assessment of compliance with values like equality, transparency, proportionality, and due process, among others.
Finally, India should foster the concept of digital due process. Digital due process would require that algorithmic governance remain consistent with principles of natural justice. Citizens should possess notice, explanation, review mechanisms, appeal rights, and procedural transparency. Constitutionalism in the digital age must ensure that technological systems remain subordinate to democratic values rather than replacing them.
In time, India may have to create its own model of “algorithmic constitutionalism,” whereby AI used by the state is treated like any other exercise of public power that must pass the test of constitutionalism. It will mean that algorithms must meet standards similar to those that the constitution imposes on public institutions.
Conclusion
The constitutional problem for the coming century will not be about the concentration of power in institutions, but about how power is distributed in these new unseen algorithmic systems. It is undeniable that the artificial intelligence system has great administrative benefits. It is efficient, speedy, and expands the ability of governance. But constitutional democracies cannot assess the efficacy of their governance merely on the grounds of efficiency. They gain their legitimacy not only in the outcomes of their activities, but also in the way they use their power. In this new era of algorithmic governance, India needs to rethink its constitutionalism. As the former sought to humanize governance through accountability and reason, algorithmic governance may well produce efficiency at the cost of legitimacy and accountability. Not because we should fear the domination of machines over the governments, but because the machines might enable the government to become unaccountable. Indian constitutionalism of the twenty-first century will depend upon the effectiveness of its democratic institutions in taming the power of algorithms before they start governing us in ways we do not even understand.
If constitutional democracies fail to regulate algorithmic power today, future citizens may lose freedoms not through visible authoritarianism, but through invisible computation. Ultimately, the central constitutional question of the digital age might boil down to one: who truly governs-the Constitution, or the algorithm?