INTRODUCTION
Aadhaar, digitalisation, Direct Bank Transfers, and social security schemes such as PMJAY and Swasthya Saathi will bring a radical change to the welfare system in India, thereby encouraging transparency and efficiency. These algorithmic systems not only sustain the creation of technology but also serve as a mechanism for repairing the doubling and corruption, resulting in an even greater problem1. That issue is a profile inclination among the poor. Now, data mining, biometric analysis, and eligibility algorithms quietly remake poverty into a measurement and management, by unseen systems that host, behind closed doors, those who may receive help and those who may not2.
This paper excavates a fundamental problem: how algorithmic targeting of welfare in India may cause discrimination, such as some families never receiving the benefits they are entitled to. Even though they are presented as neutral and scientific, algorithms do not exist without a set of assumptions, labels and priorities created by policymakers and engineers. These design decisions in welfare schemes, in many cases, tend to strengthen the old social order by striking the most marginalised, such as urban migrants, female-headed families, SC and ST and the urban poor. Blocking people from receiving rations or deleting patient names under Ayushman Bharat indicates the presence of digital inequality3. This question is important because it lies at the intersection of distributive justice, constitutional rights, and digital privacy. Reason: Consequently, the principal reason persuading the court was that the individual named in this case (Puttaswamy) was notorious and had frequently altered the statute to compel the government to approve his behaviour, which constitutes an offence.
In Justice K. S. Puttaswamy (Retd.) case. v. Union of India, right to privacy was hence a fundamental right and should therefore be aware of the risks of profiling. The welfare system was profiled even after the Aadhaar judgment. This created a conflict between the government’s desire to be efficient and the Constitution’s protection of privacy. Examples include the UP Universal Credit System, the world’s largest4. The US fraud detection system and the global move toward algorithmic welfare indicate that India is not alone in mucking with these problems5. The case of India is quite unique due to the sheer magnitude of its welfare, the abject vulnerability of its poorest populace, and the fact that online governance is shrouded in obscurity, making it difficult to read.
In Part I, following the introduction, the rise of the algorithmic governance in Indian welfare programs, the rationale behind it, and why it has been embraced in India, down to the clatter of computers in government bureaucracies. Part II goes deeper into the realm of doctrine and policy, exploring principal tenets of the constitution, tracing the lineages of legislative structures, and the most significant decisions taken by the court, such as one that has transformed debate in a single night. Part III contains case studies of welfare exclusion and discriminatory outcomes, and draws comparisons with how other jurisdictions address similar problems. Part IV examines the comparative and global dimensions. Part V provides the impact of algorithmic systems on stakeholders, including urban poor families, migrants, and other marginalised groups who experience the rough sting of exclusion. Part VI provides cutting-edge commentaries and practical policy concepts that call for reconsidering welfare targeting through a rights-oriented approach with an unambiguous, consistent focus and part VII provides critical reflections and recommendations related to algorithmic discrimination in Welfare, targeting in India.
The paper concludes on a note of caution: the absence of transparency, actual control, and means for people to influence decisions may result in algorithmic profiling of the poor, undermining the welfare state and depriving people of the constitutional aim of equality and dignity.
BACKGROUND: RISE OF ALGORITHMIC GOVERNANCE IN WELFARE
The two issues that have bedevilled welfare programs in India are how to ascertain who really needs help and how to ensure that the assistance reaches them without finding its way into the cracks. From ration cards of the 1940s, battered at the sides, to the current biometric databases, the state has struggled to include everyone and maintain a smooth operation of the system6. During the 1990s, with economic liberalisation, the need to restrain the budget was no longer taken lightly, and welfare spending became the focus of scrutiny, with the phrase ‘leakages’ becoming entrenched as a policy concern. It was the promise of technological solutions that began to spread there, like water on dry kindling7. The Aadhaar project, the initiative of the UIDAI introduced in 2009, became a turning point. Sold as a hosted voluntary ID scheme to enable people to have greater control, Aadhaar soon became the centre of welfare provision, one that connects well with Direct Benefit Transfers. Using Aadhaar, subsidies on food, fuel, and fertilisers were linked to eliminate ghost beneficiaries and ensure that public money reached living people, not the names of dead people on the dusty register. Subsequently, the reasoning behind the Aadhaar system diffused in healthcare, education, pensions and even housing. The world of welfare changed, with local paper records forgotten and replaced by fingerprint scans and data tracking. This technology transformation did not end with Aadhaar; it extended to other areas, including bank logins and fingerprint scanners at railway stations8.
In 2011, the Socio-Economic Caste Census, developed to capture patterns of deprivation, was digitised and superimposed on benefit data archives, with its columns of names and numbers cooking on YouTube monitors. Each household was ranked based on indicators of need, and an algorithm then provided lists of those eligible to receive welfare and those who were not. With time, algorithmic scoring evolved beyond merely stating who people were and sought to incorporate them into the strategy of live beneficiary lists, in which eligibility was re-examined and occasionally updated as additional information flooded in. About efficiency concealed the cracks in the model’s design, as paint on a crooked beam9.
To begin with, India has a considerable number of poor who are out of reach of certain credible data systems; they may lack a permanent address, official work documents, and a permanent footprint beyond a phone bill. For several members of the urban poor community, particularly migrant workers, linking their Aadhaar to their ration cards or health care benefits is not easy, and this is aggravated by frequent relocation and congested, ramshackle rooms. Such biometric authentication glitches enhance the threat, such as a scanner reading a fingerprint incorrectly. Studies of the Public Distribution System reveal that millions of people have been deprived of their food rations due to fingerprint incompatibility or network glitches, and in other cases have been refused their food despite sacks of grains being right before their faces. Despite these gaps, the state continued to expand its system of algorithmic welfare, driven by two key forces10.
International organisations such as the World Bank advocated digital welfare in the name of good governance reforms, making India a laboratory for large-scale digital ID systems. Second, political leaders back home positioned digital governance as modern and clean, so that it was free of corruption and could deliver welfare more easily, such as benefits on a single tap, which made the population receptive to profiling systems11. This change totally transformed the meaning of welfare, as though the ground had been pulled up and put down on other grounds. No longer were the poor seen as citizens with rights and a demand for state assistance, but as items on a cold, flickering screen, to be put through spreadsheets and cyber-spreadsheets and so listed. Eligibility for welfare was no longer based on social/economic need, but on meeting criteria for steering the algorithms, such as clicking the appropriate boxes, meeting deadlines, and adhering to all the rules. The Algorithmic governance of welfare presents a genuine opportunity, but it has a sharp cutting-edge of danger. It accelerated the acquisition of benefits in certain people and ensnared new levels of marginality and discrimination in others12. By viewing poverty as a statistical problem rather than an embedded problem of inequality, India puts its welfare system at risk of eroding the equity that it is supposed to bring. There is a need to carefully consider laws and policies to determine whether this type of algorithmic profiling is compatible with constitutional rights and does not violate human dignity13.
DOCTRINAL & POLICY ANALYSIS
In order to implement algorithmic welfare, we must know its constitutional position, its statutory laws and data governance regulations:
A. Constitutional Foundations
(a) Right to Equality (Articles 14–15)
The application of algorithms to determine who receives welfare raises serious concerns in light of Article 14 of the Constitution, which ensures equality before the law and prevents the state from governing arbitrarily. Not only is shutting out people of Scheduled Castes or Scheduled Tribes or in other minority groups an algorithmic discrimination error, but it may also become illegal discrimination. The high court has repeatedly held that any state action must survive the twin tests of a clearly intelligible difference and a rational connection14. Algorithms are not typically going to pass when they close people out for reasons that are unknown or even just random. This is also prohibited in Article 15(1) by caste based discriminatory practice, which is even more significant when databases of SECC and welfare embed caste labels15.
(b) Right to life and dignity
In Olga Tellis v., the Court made its decision with sharp reasoning. Bombay Municipality Corporation recognised that the right to live is part and parcel of the right to earn a living. False profiling deprives a person of food, pension or healthcare: when algorithms kill a person, it punches Article 21. Also in 2017, the Supreme Court held that human dignity is too closely intertwined with privacy as a fundamental right. The lack of open disclosure and adequate safeguards in the application of algorithmic profiling by welfare agencies, in the absence of clear consent, leads to people losing their privacy and their right to make their own decisions16.
(c) The third is Directive Principles and welfare jurisprudence, the opposite.
The Directive Principles of State Policy (Articles 38, 39, 41, and 47) of the Indian Constitution urge the state to facilitate or promote welfare, secure livelihoods, and ensure social justice, such as ensuring that all families have food on the table and comply with the law. These principles cannot be adjudicated in the court of law, but they have been used to determine how judges can appreciate welfare entitlements as true rights17. Consider PUCL v UOI, as since 2001, the Union of India has recognised food security as a fundamental right to which the people of the United Arab Emirates may pursue judicial redress. In cases where an inaccurate algorithm locks people out, such as the refusal of a family to apply to have a house, it conflicts with the Constitution’s emphasis on the welfare of the people18.
B. Legal and Regulatory Environment
(a) Aadhaar (Targeted Delivery of Financial and other subsidies and benefits and services) Act, 2016
The Aadhaar Act provides a legal basis for the application of biometrics in welfare services. In 2018, the Supreme Court, in its Puttaswamy (Aadhaar) case, supported Aadhaar for welfare benefits but prohibited its use by a business entity, thereby emphasising that its use should be proportionate. As a matter of fact, Aadhaar seeding and authentication cannot be avoided, and individuals do encounter them without any effective system to complain about them. It erodes statutory protection and undermines the proportionality test the Court has established19.
(b) Data Protection and Digital Governance:
Protecting customer data and the company must be safeguarded and monitored to prevent security breaches, potential information attacks, and employee fraud. The protection and administration of data kept on the customers and the organisation should be secured and monitored in order to censure any protection incursions, possible assault on information, and fraud in the company by workers20.
(c) Legislative restrictions
India’s National Food Security Act, 2013 (NFSA), and the Mahatma Gandhi National Rural Employment Guarantee Act, 2005 (MGNREGA), entrench the rights to employment and to food, even guaranteeing a sachet of rice per day. The NFSA ensures that two-thirds of the nation can obtain food grains, such as rice or wheat, at subsidised prices in burlap sacks from the market. An algorithm blocking someone from receiving rations is not just a bureaucratic slip; it is the law. Similarly, the fact that Aadhaar malfunctions results in an individual being excluded from the MGNREGA job list is a straightforward abuse of his or her rights under the law21.
C. Court Interventions and Case Law
These instances have laid a prominent way to the rights of welfare, but there are loopholes due to the absence of machine learning based bias:
• PUCL v. Union of India (Right to Food case): Article 21 had its scope expanded and encompassed food security that now has entitlements which cannot be denied or risk being considered as a violation22.
• K.S. Puttaswamy v. Union of India 2017 Privacy was a fundamental right, and therefore, it implied algorithmic profiling23.
• K.S. Puttaswamy v. Union of India 2018 Proportionality and some limited exclusions were required, which retained Aadhaar as a welfare provision24.
• Swaraj Abhiyan v. Union of India (2016), declared that the application of the National Food Security Act 2013 and that a state was obliged to offer welfare in the event of distress without any form of administrative interference25.
D. Doctrinal Challenges
There are two main issues of dogma:
1. Openness and Transparency: Welfare judgments are frequently based on secret algorithms, such as those in Telangana, where a screen flickers with information to influence the benefits people receive. In cases where information is held behind closed doors, individuals would strike walls, attempting to question those in power in court26.
2. Structural Discrimination: It might entrap structural injustices in place, such as dust settling in the cracks of old stone, based on SECC caste and deprivation data. The courts will need to develop new provisions of algorithmic disparate impact, such as the indirect discrimination discussed in Article 1527.
Policy Implications
The doctrinal critical analysis shows that the world of algorithmic welfare presents a sharp contrast between the promise made in the constitution and its actual implementation. Technology in itself is not unconstitutional, but in the absence of checks such as transparent audits, an appeal mechanism, and actual accountability, the system conflicts with the rights-oriented welfare model in India28.
CASE STUDIES
In this section, we will examine significant state initiatives in an attempt to demonstrate how the idea of algorithmic welfare targeting, as it is currently being applied in practice, has transformed the process of identifying beneficiaries, all the way up to the point where the digital form notices the name on the screen and informs all the associated agencies. These case studies demonstrate how algorithmic systems can streamline workflows according to schedule while still isolating individuals in ways no one has anticipated.
(a) Telangana: Entity Resolution and Samagra Vedika.
Telangana Samagra Vedika is termed by some the most hi-tech experiment in data-driven governance in India, which kills millions of data particles flickering across screens and walls with every decision-making process. Beginning with an initiative launched in 2018, the project consolidates over 25 departmental databases, including welfare, tax, property records, and household information, into a single, easy-to-use platform with an easy-to-switch-on switch. It is supposed to be a reference to a single truth in governance, and officials can easily verify who is eligible for welfare under various programs. Among them is entity resolution, which is carried out in collaboration with the private tech company Posidex Technologies29. Machine learning algorithms align individuals and homes across various data sets, identifying redundancies or duplicate records that do not exist. Paperwise is supposed to prevent fraud and seal any loopholes, such as preventing a drip before it can wet the floor. However, in fact, it has been audited and documented that persons are being turned away everywhere30.
Exclusion Episodes: Thousands of pensioners and ration cardholders, long on the list, have been abruptly removed after Samagra sounded their names on the list as dupes, and now have their tattered cards in their hands that no longer count. The mismatch, which occurred most often in urban poor and migrant families, was frequently caused by minor spelling errors, lack of movement, or a lack of standardised addresses31.
Secret algorithms can deny benefits without explanation, and victims have no opportunity to complain or punish anyone. Samagra Vedika shows how large-scale automation, paired with hidden vendor tech, can quietly stack the system against vulnerable people even as it parades itself as “scientific” governance, polished and humming like a well-oiled machine32.
(b) Aadhaar-Backed Public Distribution System (PDS) Jharkhand.
Jharkhand lies at the centre of any Aadhaar-dependent welfare, with both local officials and villagers taking on each side of the villagers under the fans, which move the humid air. The state has also linked Aadhaar to the Public Distribution System (PDS) to eliminate ghost ration cards, and now people verify their identity with a simple fingerprint scan on a small biometric device at the ration shop33.
• Authentication Failures: Researchers have found that, due to worn-out fingerprints, which are common amongst those manual labourers, who have their hands scratched each day and which are patchy, the network signals tend to block legitimate receivers of their monthly food grains. In 201718, governmental audits revealed that, in the event of authentication failures, hundreds of thousands of families received their rations at home empty-handed34.
• Tragic Consequences: Civil society documented the cases of starvation deaths that were directly attributed to a failure of the PDS associated with Aadhaar, similar to Santoshi Kumari, an 11-year-old girl, who was reported dead in 2017 after her family was denied grain due to the inability to seed their ration card35.
Systemic Impact: The officials dismissed the exclusions as minor technical malfunctions; however, when combined, they struck hard, leaving some of the poorest Dalit and Adivasi families in the state scrambling to put food on the table. The case illustrates how a very basic algorithm, say, a biometric fingerprint check, could decide who to eat and not to eat when it is associated with something as relevant as food security.
(c) Tamil Nadu: Dynamic Targeting and Family Database.
The implementation of welfare schemes in Tamil Nadu, particularly the PDS, which is said to cover all people, has been widely admired. More recently, the state has shifted to more dynamic forms of welfare targeting, relying on a Family Database System (FDS) that tracks the changing needs of families, such as light following movement. The FDS links multiple welfare programs to a single household ID, enabling a person to easily confirm eligibility for pensions, scholarships, or subsidies36.
Dynamic Exclusion: Authorities can rapidly change a household’s eligibility when new financial betterment signals are present, such as a car purchase, property registration, or increased power consumption. Nevertheless, these signs are crude surrogates deprived of actual signs of vulnerability, such as the shaking in someone’s voice when one is scared. A used car is still capable of getting you disqualified, even when the family is barely making ends meet below the poverty line.
• Less Openness: Individuals will typically discover that they have been dropped off the list at the counter like a ration counter, school desk, or hospital window,
• Private Vendor Role: Once again, similar to Telangana, system design and data integration are assisted by private IT companies, which raises the question of who is to blame once a glitch slips in. The experiment in Tamil Nadu shows a transition from rights-based universalism to conditional algorithmic targeting, which could negatively affect inclusiveness, a characteristic that has previously characterised its welfare model.
Comparative Lessons of Indian States.
In all these case studies, there is a certain theme:
I. Risk of high exclusion: Algorithms tend to leave out individuals who are in need of assistance, be it entity resolution in Telangana, which is checking fingerprints in Jharkhand, and targeting dynamically in Tamil Nadu, and some people are forced to leave with no money37.
II. Opacity and Vendor Dependence: Reliance on confidential technology vendors obstructs the procedure and erases the view of transparency and accountability, as frost does a window. The system’s algorithm remains a secret in private systems38.
III. Caste and Social Disadvantage: Dalits, Adivasis and Muslims may endure a disproportionately large number of unjust exclusions, a real testament to how algorithmic reinforcement can be so insidious.
IV. Weak Grievance Systems: The individuals who are marginalised have seldom identified actual means by which they can dispense justice, and, as a result, the suffering accumulated silently and, in some cases, that silence was a life worth losing. These instances at the state level demonstrate that algorithmic welfare targeting, despite its contemporary pretensions, tends to sharpen the exclusion edge-cutting tools, notwithstanding its promoted rationality. They emphasise that we so desperately require our own audits, clarity in transparency standards, and stricter legal safeguards that welfare technology really reinforces ideals of equality and dignity expressed in the constitution39.
COMPARATIVE/GLOBAL DIMENSIONS
India is not the only country with a difficult issue of algorithmic bias in welfare programs; it is a fight that reverberates everywhere where data determines who would be assisted and who would wait. Governments across the globe are resorting to algorithmic applications to disperse social benefits, as they are enticed by quicker service, enhanced fraud detection, and reduced costs. Nevertheless, a glance at the practical examples reveals that the systems have a tendency to intensify rather than address structural inequities. A close examination of case studies in the U.S., Europe, and some developing countries throws light on the risks and the safeguards India may require40.
(i) USA: Relativism and Due Process of the Law in Automated Decision-Making.
In the US, algorithmic tools are currently being implemented across numerous state welfare initiatives to accelerate eligibility checks and fraud detection. The most infamous example is the Michigan system, MiDAS, which automatically checks individuals for fraud by verifying their employment status. The system threw clinkers: 93 per cent of its notifications were erroneous, and even accused tens of thousands of innocent recipients of extortion, costing them money, trust, and even a day in court. Subsequent legal tussles revealed the inherent inhumanity of automated systems: they trample due process, lack real human supervision, give no warning, and have no appeal process41. In the case of India, it demonstrates how unsafe it is to hand over algorithms too much when determining who receives welfare, and one bug in the code can leave a family without anything on the table. Devoid of solid structures to question errors or challenge biased results, individuals falsely excluded from Aadhaar-based targeting or a digital malfunction could face the same harm, such as receiving their monthly ration card without any obvious means of rectification42. We observe in the U. S. that fair process is being undermined by hidden technology, where individuals are unable to comprehend what motivates an algorithm to make a decision and thus are unable to challenge it; it is akin to arguing with a black box and humming under glass.
(ii) Europe: Human Rights and the Syrian Case in the Netherlands.
The trend among European courts and regulators is now shifting toward a right-based approach to the governance of algorithms, grounded in the EU Charter of Fundamental Rights and the GDPR. In 2020, one of the milestones was a Dutch court’s decision to strike down the System Risk Indication (SyRI), a data-driven tool that was previously used to identify welfare fraud and tally numbers. It turned out that the court said that SyRI violated the principle of non-discrimination and the right to privacy as it targeted low-income and immigrant areas43. The experience of the Dutch demonstrates that welfare algorithms are not only mechanisms of technology; they are also arenas where the rights of people are determined, as tangible as a pile of case papers on a table. The SyRI case in India demonstrates that the Indian Constitution does not view algorithmic targeting as a mere administrative failure and protects equality under Article 14 and dignity under Article 21. When courts begin to consider opaque algorithms as a basis for infringing on rights, they may be able to remake accountability in the Indian welfare system44.
(iii) Developing Nations: Kenya and Brazil.
Third-world nations experimenting with digital welfare states echo the journey the Indian nation is undergoing, as a screen saver on a handheld phone. In Kenya, the purpose of biometric systems was to automate the delivery of assistance to refugees, though most individuals were excluded because their fingerprints or eye scans were not detected, particularly manual workers with rough, calloused hands that the sensors could not detect. There is a similar response elsewhere, including in India, where Aadhaar checks have frequently blocked the operations of older people, disabled individuals, and day labourers who are required to stand in dust-filled lines to claim their salaries45. The conditional cash transfer program known as Bolsa Familia in Brazil, with its less frequent biometric checks, has been criticised for making mistakes caused by algorithmic choices calculated based on data-driven eligibility criteria that have the unfortunate effect of marginalising the poor and excluding those who are most in need of welfare benefits due to their family size or location46. In regions of high inequality where data barely has enough data to support its machines, algorithms usually miss the mark, driving the poor out of the system and leaving those who need welfare assistance the most on the house. When applied to the situation in India, given the reference to the same harsh social and economic conditions, the real problem is not simply faulty code, it is the deep structural imbalance entrenched in the data doing the writing, like fissures through stale cement47.
(iv) Lessons and Comparative Insights.
An interjurisdictional scale, by comparison, starts with pointing out three themes that have emerged across a variety of jurisdictions:
A. Opacity and Lack of Accountability: Welfare algorithms are sometimes treated as sealed black boxes in Michigan, the Netherlands or India, which means that a beneficiary can no longer have an explanation or even an actual opportunity to appeal against an unjust ruling48.
B. Oversized Impact on Vulnerable populations: All across the globe, the automated processes disproportionately strike upon marginalised populations such as immigrants in Europe, racial minorities in the U.S., informal workers in India, and refugees in Kenya, languishing in an oppressive camp in search of a fair opportunity. Welfare systems intended to assist people are often rendered ineffective.
C. Judicial and` Regulatory Interventions: Shared experiences include how the courts and data protection legislation come to bear as a means of reining in runaway algorithms, the mind thinks of a judge tapping the gavel to get proceedings back on track. The SyRI ruling of the Dutch court, the boundaries of automated decisions set out by the GDPR, and lawsuits on due process in the US are examples of how the law can intervene to reduce the bias of algorithms. India can also intervene through its new DPDPA (2023) and a well-developed system of constitutional rights49.
Implications for India
The overreliance of India on Aadhaar-based systems places it at the heart of discussions on the role of algorithms in determining welfare across the world. As the experience shows, new technology, however clever it may be, cannot take the place of the clear, stable security of legal safeguards. To focus on welfare, keen information gathering, explicit regulation, and a strong mechanism for decision redress are necessary, and a practical appreciation of the underlying structural imbalances in action is essential. India can develop welfare initiatives that would draw more people into the system rather than leave them out by examining what has worked and what has not spread around the globe50.
STAKEHOLDER IMPACT
Focusing welfare in India through the application of algorithms concerns not only people directly affected but also those on food rations, whistleblowing offices, technology firms, and the rest of civil society. Mapping them out allows us to see the unpunished second-order consequences of algorithmic discrimination, and it is a decent reminder that those in the welfare system should be made specific to achieve a balance between justice and speed51.
A. Marginalised Communities and Beneficiaries.
Algorithms in the targeting, which are most destructive against the poor, desperate, who are on welfare, do so because they retrieve it up to the last grain of rice. The linking of Aadhaar IDs to biometrics and the conduct of data-intensive eligibility checks have marginalised a significant number of people. Research has shown that a significant portion of beneficiaries have been refused rations, pensions, or healthcare services due to fingerprint mismatches, network failures, or inability to access the old databases52. The disabled and migrant workers, as well as old patients whose fingerprints have been worn away, cannot easily identify themselves. This is by no means a technical loophole. Being made to starve courtesy of the Public Distribution System or withholding someone of the right to good healthcare courtesy of Ayushman Bharat, survival and dignity are going down the drain just as fast as a bowl of rice that is not eaten. Mostly, the targeted, tagged, or even closed down are ironically the people who are dependent on welfare. The use of algorithms only further increases inequality, most notably among people who are already marginalised, as it happens when a door is closed in their face53.
B. Assurance of reliance on Algorithmic models.
Algorithms targeting the state are a cheaper, faster way to operate and are more effective at fraud identification. The government computerises checks and money transfers to minimise leakage and ensure aid reaches the right hands, not the wrong ones, like a single mother waiting at her kitchen table until the deposit comes. Nonetheless, depending on the algorithmic models, responsibility shifts elsewhere as well. When it fails, government bodies call it a technical glitch, and politics is removed from anything with extremely real political effects. Other than that, algorithmic governance increases the state’s ability to track its own people, including the ability to track a phone even in the middle of the night54. The welfare databases are currently connecting to voter lists, medical records, and even bank records, creating information centres that are the central points where it is hard to tell where public help ends and surveillance begins. This is in actual danger of proper democratic accountability and makes citizens a set of data that is under continuous scrutiny.
C. Technology Firms and Privates.
The players in the biometric solution and data analytics service, individual technology actors, and vendors of Aadhaar-related services are at the centre of targeting welfare needs (such as mapping city streets and sending relief trucks). In India, digital welfare systems work based on their code, image-checking programs, and hardware and software, as the code validates the ID against servers operating in Indian government data centres. However, the advent of the private sector raises concerns about questionable decision-making, a profit-driven model, and the absence of anyone publicly bearing responsibility for the consequences55. The fact that many algorithmic models are guarded by a key behind their walls exposes them to threats to company secrets when they transform into large organisations. It does not allow citizens, activists, and even judges to work down to the composition of those decisions, including tracing the submerged ink of a signature to the hand that wrote it. Technology companies are cashing in on government decisions and massive data-gathering, whilst the poor risk being misplaced or discriminated against, like a form being taken off a screen before it is read. Welfare systems privatised to companies tend to interweave business interests and lose accountability56.
D. Referral to the Civil Society, Activists and Legal Advocates.
The civil society organisations help them remain vigilant about power, document instances of sidelining, file cases against them in court on behalf of the populace, and mobilise affected communities against algorithmic prejudices. In India, activists have demanded justice for the starvation Aadhaar authentication deaths, which have shown that a simple malfunction can cause an unserved meal and the taking of a life. The attorneys have been defending their constitutional rights, equality, dignity, and privacy, seeking to subject the exclusionary welfare policies to the Supreme Court and the High Courts, crying foul over the principles that disqualify certain families from receiving cold. However, algorithmic bias, such as a job-screening application automatically filtering out successful job applicants, is hard for civil society to oppose. Because algorithms are black boxes, it is hard to provide clear-cut evidence of systematic bias57. The state is also inclined to cast criticism as anti-technological development, to the point of calling activists anti-development. However, it is civil society that plays an important role in ensuring that welfare systems are fair, inclusive, and premised on human rights, just as a community library is accessible to everybody.
E. International Organisations and Donors.
The Indian Aadhaar program has served as a model for major world players, including the World Bank and major international donors, of the digital inclusion concept. They have left a lasting mark by establishing financial accessibility, streamlined procedures, and technology-based governance to ensure that the nation focuses on its own welfare, including the use of fingerprint scanning at local enrolment centres. However, exceptions have occurred in which this refrain of global positive reinforcement has overtaken the no-go-slow debate over rights, non-partnership, and who seems to be watching. The global community poses a critical question: how much should developmental resources be stretched, given that guardrails are not firmly in place, as in the case of implementing new drones without testing their chassis?58 In India, the applicability of international principles of digital governance could create a new wave of innovation, but, at the same time, could entrench an exclusionary approach, with international quality labels stamped on people with mahogany-coloured skin.
F. Long-Term Social Impact of the Issue.
Broadly, by rewriting the rules of the game, algorithmic welfare targeting is changing the citizen-state relationships. Benefits guaranteed by welfare that were formerly based on the principles of rights and equality on a general level have been complicated, preconditioned, information-supported and algorithmically filtered. The poor do not become rights-bearing citizens but are reduced to the subject matter measured by riskiness, such as dollar signs in an Excel sheet. The outcomes of this transformation resemble the smell of disruption of solidarity and have granted profiling the imprimatur of approval and the process of exclusion the aura of something holy, rather than something on the wiring sheet. The uncontrolled, welfare targeting has silently reinvented social ladders of the ancient, and aided caste, classes, and genders in the well-furnished nomadic outfits of neutrality to technology59.
EMPIRICAL INSIGHTS
Despite the fact that most debates about algorithmic welfare in India revolve around law, doctrine, and policy, real-world data can provide clear insight into how such systems can work in contexts such as overcrowded ration shops or rural registration centres. Non-reliant research, state analysis, and field ethnography reveal the extent of deep bias in welfare algorithms, manifesting as patterns, boundaries, and the human cost, such as a mother losing her benefits due to a flawed capital private data entry60.
A. Theimenfelder Fehler in Aadhaar-Welfare.
Most research indicates that Aadhaar biometric verification tends to lock out individuals, sometimes leaving one at a ration counter with the thumb stuck against the scanner until someone is verified. A Right to Food Campaign survey conducted in Jharkhand in 2018 found that approximately 10 per cent of ration cardholders were not receiving their food grains because the system could not verify their ration cards. The majority of these failures came to light among older adults, people with disabilities, and manual workers whose hardened, calloused fingertips were difficult to scan. Similarly, the National Economic Survey acknowledges that even with biometric systems, duplication and fraud are reduced, yet still excludes a number of individuals61. Approximately 2-4 per cent of those intended to receive welfare benefits are left out due to mismatches or outdated information caused by poor infrastructure. The size of these percentages may seem small, but when a country is talking about welfare programs that reach hundreds of millions of people, millions of people still fall outside the net, even when their lives are ruined by the misplaced swipe of an algorithm62.
B. Mortality Data and Starvation Statistical Data.
There have been at least 20 killings during 2017-2019, to which people in Jharkhand, Uttar Pradesh and Odisha have been denied food rations due to not having Aadhaar, or their names having disappeared off the rolls following so-called de-duplication drives. Even though officials tend to dispute the numbers, every case demonstrates that algorithmic exclusion can be extremely deadly. They also demonstrate that exclusion is hardly random, but rather conducted in some pattern so that those affected are usually Dalit, Adivasi, or other discriminated groups who have little to no means of pursuing justice, with their grievances going down a never-ending list of not responded papers63.
C. Forensic Biases and Errors of Inclusion, Misclassification.
Statistics indicate that exclusion errors take the headlines, but inclusion errors persist. According to a CAG report on the Public Distribution System, despite linking Aadhaar numbers, databases still contained ghost centres and duplicate names. The fact that these errors are so difficult to overcome indicates that algorithmic targeting does not have a magic bullet to save all the efforts. The only bad thing about misclassification is how it operates; those who have worked their way there are locked out by the cold, unbending bots64. To illustrate, families living in densely populated slums in urban areas have been excluded from health programs because their income documents fail to meet official requirements, yet their simple houses with tin roofs clearly show they were poor65.
D. Personal Experience and Ethnographic Data.
Ethnographic observations provide human colour to the numbers. Rajasthan and Jharkhand have revealed that beneficiaries feel the number of such algorithmic decisions is cold and pointless, like a faceless screen that judges their lives. Most individuals complain that they are fed up because they had no idea why their ration cards had been cancelled or why the fingerprint scanner was being recalcitrant. Algorithms do not exist to them as abstract systems. Interviews indicate that such a rejection sets off a muted wave across all that is to come. In cases of rationing, families are put into a debt hole, and in cases of withholding health benefits, treatment is either delayed or not received. When one has had enough refusal to enter his/her logins, faith in the welfare system breaks open as an old paint does. There are instances where beneficiaries internalise the toughness and blame themselves, to the point of further entrenching a sense of marginalisation and making them feel like someone wants to close the door on them66.
E. Statistics regarding Correspondences and Referrals.
On-hand data shows a drastic disparity in the resolution of grievances. In the states where such a system of complaints is organised, only a small portion of people utilise it, become lost in long channels and lose hope in front of some frozen, piled with papers online questionnaire. Even in Jharkhand, only a small percentage of the left out appealed for the cancellation of ration cards. According to field reports, most low-income families lack literacy, internet access, and, in some cases, even a charged cell phone to navigate the appeals process67. Exclusion after it occurs, with no actual means to solve it. This makes momentary seizures of algorithmic twitching permanent damage, the one that leaves as a lasting experience, as shutting down a light you believed was the one that is destined to flicker again.
F. Statistical Trends on a broader basis.
Compared to looking within states, the picture of algorithmic exclusion is evidently not evenly distributed; there are rougher spots, as shown in a map where the darkened areas appear more frequently in the same locations. Poor internet connectivity and large numbers of marginalised people in some states, such as Jharkhand, Bihar, and Chhattisgarh, will result in more individuals being outside the system68. In comparison, the performance of states like Kerala or Tamil Nadu is lower, suggesting that good political intent and effective governance influence how algorithmic welfare systems may actually perform.
G. Key Empirical Takeaways
1. Scale of Exclusion: It only takes minor percentage differences to result in millions of people excluded, and due to the context of life-and-death situations, extreme poverty.
2. Systematic Discrimination Patterned Discrimination: discriminatory exclusion does not happen by chance but is focused on disadvantaged communities, which include the Dalits, Adivasis, migrants and the elderly.
3. Weak Redressal: This refers to the situation whereby an appeal mechanism is not available, and the majority of the exclusions are not corrected.
4. Global Resonance: HR in India matches global results with a bias in favour of systemic and not isolated malpractices. All these empirical findings prove that algorithmic welfare targeting is not a neutral instrument of administration and can alter human lives in real and unequal ways. They play a significant role in providing an evidentiary foundation for doctrinal reasoning on inequality, dignity, and the rights to food and health in India69.
CRITICAL REFLECTIONS AND RECOMMENDATIONS.
The above segments indicate that algorithmic welfare targeting in India is not technologically neutral and socially benign. Rather, it represents institutional injustices, exacerbates statistical exclusions, and deepens the lack of transparency in government. Drawing on comparative models and normative principles, this section critically assesses the major gaps in the existing framework and provides recommendations for reform.
A. Critical Reflections
I. The Fallacy of Neutrality
Often, people refer to algorithmic systems as neutral tools that can eliminate human error and even apparent hints of corruption. These systems, as the evidence demonstrates, are highly driven by the datasets on which they are based, particularly the SECC, Aadhaar records, and other related administrative files, and some of them are as accurate as household counts or year of birth. These data bear the burden of the past, such as biased documents and misplaced tags. Welfare programs based on automation can perpetuate old biases rather than remedy them70.
II. Exclusion by Design.
The greatest impact of Design Bias of algorithmic models is experienced by the members of the marginalised community who experience discriminatory access to food, health, and dignity. These systems are more concerned with cost reduction than with creating space for inclusivity.
III. Lack of Opaqueness and Dissatisfaction with Accountability.
Private vendors such as Posidex do not allow others to use their proprietary software due to intellectual property laws. The beneficiaries, activists, and even bureaucrats are at times unable to explain why someone became an exception, even when the paperwork is in front of them. Such obscurity undermines procedural fairness and diminishes democratic accountability, as deals are made when they have been decided by visible administrative judgment, only to become decisions in invisible systems that no one can understand71.
IV. Weak Redressal Mechanisms.
When individuals are unable to make complaints and seek assistance with ease, marginalisation sinks its roots deeper. There are not only no appeals, but they are either buried under red tape or totally broken. People require literacy, internet connections and patience to cut through red tape to use the digital grievance. Therefore, if the algorithm malfunctions, the harm may be lasting and lock individuals out of resources72.
IV. Mismatch between Rhetoric and Agreements.
According to ethnographic data, welfare systems nevertheless do not appear to be very close or friendly to residents of marginalised areas. The basis for this technocratic optimism about welfare is anything but the hard realities people face in their day-to-day lives73.
B. Recommendations
I. Enhancing the Principles of Data Justice.
Any actual reform must incorporate data justice into the administration of welfare, such as embedding it into every policy line.
• The accuracy of information should be ensured: It implies that SECC and Aadhaar-related datasets should be regularly updated and thoroughly checked, such as re-checking a name or address. Non-discrimination Safeguards: Welfare algorithms are to be audited regularly for disparate impact based on caste, gender, region, and disability. The welfare algorithms must be regularly checked to identify any form of discrimination, whether based on caste, gender, region, or disability.
• Right to Explanation: The idea of the right to explanations should follow the example of the EU General Data Protection Regulation (GDPR), as beneficiaries relying on it need a right to understand why they were included or tracked down and the reasons for that decision.
II. Institutionalisation of the Algorithmic Audits.
There must be a legal structure to enforce an algorithmic audit as an obligatory aspect of a process that institutions perform, such as locking the doors at the end of the day. Based on the Netherlands and Canadian models, India ought to establish the Welfare Algorithm Oversight Authority, which:
• Conducts pre-deployment testing on welfare algorithms that satisfy, e.g. how they behave on an income data set of a family in the real world.
• Issue reports of rates of exclusion and inclusion errors. Review the vendors’ contracts for transparency and accountability.
III. Finding a Balance between Efficiency and Rights.
Go about the policy with the intent of attracting people, not locking them. The world experience demonstrates that welfare systems, which are constructed based on the principle of the inclusion-first approach, being ready to permit a certain degree of overlap, are more likely to be fairer and more successful in decreasing levels of poverty, such as making the door wide enough to accommodate everyone who really requires it. India must make an intentional change in welfare distribution to reduce improper eliminations.
IV. Increasing Disgruntlement Redressal.
Complaint management systems with many well-defined layers are necessary. Recommendations include selecting independent grievance officers in every district to override algorithmic decisions. Online and offline procedures should be used to ensure accessibility. Redressal of the welfare appeals is to be performed in a time-bound manner.
V. Co-Design and Community Participation.
The creation of an algorithmic welfare system cannot be built by a small, obscure group of non-democratic bureaucrats and technology sellers. These systems that beneficiaries, regional governments and community organisations form when they design at the location embody life on the ground. Different checks are required before any new digital welfare is implemented through reviews, social checks, Gram Sabha reviews, and participatory impact checks, among others.
VI. Lessons in Comparison and Reconstruction.
By comparing with other countries around the world, India can learn important lessons, such as in the United Kingdom, where the algorithms behind Universal Credit were opaque regarding eligibility. The welfare of Syrians in the Netherlands was ruled on because it infringed on the right to privacy and equality. In Brazil, community oversight is implemented through the Bolsa Família programme.
VII. Future of Rights-based Legislation.
Eventually, long-term reform will be attained when we inscribe algorithmic accountability into the law itself, in black and white, and not merely lipstick lip-reading. The Digital Personal Data Protection Act of 2023 in India is insufficient; it focuses on privacy but does not address discrimination. An identified Algorithmic Accountability Act must categorically recognise algorithmic exclusion as a legal injury and acknowledge that being locked out of a job portal by bad code is like banging your head against the wall.
• Requirement of the disclosure of datasets, logic and error rates.
• Guarantee of rights, the right to explain, to appeal, and to remedy, which shall be enforced.
• Demand open display of datasets, logic behind it and rate of error.
C. August 2019. Future Research/policy questions.
1. Efficiency and protection of rights: Can the algorithms do this?
2. Is it possible to audit algorithms when the state controls the political economy through vendor contracts based on the state?
3. How are courts addressing algorithmic exclusion? What conceptualisations should courts adopt in addressing algorithmic exclusion, in terms of equality, arbitrariness, or the socio-economic rights?
4. What is the global mark on the algorithmic governance of welfare between international human rights standards?
5. Is it possible to nationally represent the diversity of India using participatory design practices?
CONCLUSION
In this paper, we have discussed how algorithmic welfare targeting in India, based on SECC data, Aadhaar-linked systems, and machine learning, has entirely rebalanced the very concept of social protection, including the mechanism for distributing ration cards to households. These systems, rather than delivering the promised efficiency and impartiality, have intensified structural inequities, excluded people and placed greater demands on prominent societies. This is the crux of it all: technology in welfare governance itself is never neutral, it is formed through and through by the datasets it attracts, the institutions and their construction, and the political choices that shape it.
Analysis began by contextualising welfare digitalisation in line with India’s constitutional pledges, such as equality, dignity, and socio-economic rights. An incisive examination of the doctrine showed that Dalits, Adivasis, migrants, and older people were among the most severely affected groups in algorithmic exclusions, as they hit a vital right and undermined the state’s obligation to provide food, health, and social security. Looking at the examples of the Netherlands, Brazil, and the UK, it is made clear that India is not the only country to struggle with, because, in the entire world, welfare technology is struggling with the process of trying to figure out how to be held to account when no one is supposed to. India has not yet erected an effective fence against the risks of algorithmic bias, unlike the real guards that judges or lawmakers have already created. These normative claims were supported by real-life evidence. Polls, inspections, and field research indicate that small fractions of marginalisation can marginalise millions of people, even though they are small in number. Multiple deaths of starvation and wrongful classifications by welfare technologies illustrate the extent to which social hierarchies currently exist everywhere around us.
A closer examination found four essential gaps: the myth that algorithms are neutral, the belief that efficiency must be first, the notion that decision-making must be kept secret, and the claim that grievance systems are so weak that they cannot even stand on their own in practice. In reply, the paper collected a change plan: weave justice with data into all processes, build audit into the independent cycles of algorithms, design with inclusion at the centre, make the grieving processes responsive, invite people into the design meetings and make it based on rights-founded law that will become solid as stone. A combination of these actions prepares a proper route to align welfare technologies with constitutional and human rights commitments. Ultimately, the argument is obvious: digital welfare reforms cannot be based solely on efficiency claims and cost savings, however neat the spreadsheets may appear. They must be counterbalanced by other higher values, such as justice, inclusion, and human dignity. Unless India starts reconsidering its welfare algorithms with a rights-first approach, digitalisation may become even more discriminatory and exclusionary. The poor are people with rights, not just figures on the computer monitor, in a just digital-era welfare state, and their rights should not be hidden or concealed, but openly defended and enforced.
*****
Footnotes
1. Vishnu IAS Academy, How Digital Technologies Are Transforming India’s Welfare State: Aadhaar, DBT & CPGRAMS, VISHNU IAS ACADEMY (Feb. 11, 2026), https://vishnuias.com/digital-technologies-india-welfare-state/.
2. Mark West, UNESCO, An Ed-Tech Tragedy? Educational Technologies and School Closures in the Time of COVID-19 (2023).
3. Etienne Lwamba et al., Strengthening Women’s Empowerment and Gender Equality in Fragile Contexts Towards Peaceful and Inclusive Societies: A Systematic Review and Meta‐analysis, 18 Campbell Systematic Revs. e1214 (2022).
4. Justice K.S. Puttaswamy (Retd.) v. Union of India, (2019) 1 SCC 1.
5. Priyanka Bhattacharya, A Perspective on Fairness in Artificial Intelligence (May 22, 2022) (unpublished manuscript), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4116921 .
6. The Performance of Welfare Schemes That Are Implemented for Vulnerable Sections Is Not So Effective Due to the Absence of Their Awareness and Active Involvement at All Stages of the Policy Process – Di, CIVILSDaily (Jan. 10, 2026), https://www.civilsdaily.com/mains/the-performance-of-welfare-schemes-that-are-implemented-for-vulnerable-sections-is-not-so-effective-due-to-the-absence-of-their-awareness-and-active-involvement-at-all-stages-of-the-policy-process-di/.
7. U.N. Conference on Trade and Dev., Trade and Development Report, 2015: Making the International Financial Architecture Work for Development, U.N. Doc. UNCTAD/TDR/2015 (Oct. 6, 2015).
8. Press Info. Bureau, Ministry of Elecs. & IT, Gov’t of India, Aadhaar: A Unique Identity for the People (Oct. 24, 2024, 8:50 PM), https://pib.gov.in/PressReleasePage.aspx?PRID=2067940.
9. Prabhat Patnaik, SECC Excludes More Than It Includes Amidst High Rural Distress, PEOPLE’S DEMOCRACY (Aug. 9, 2015), https://peoplesdemocracy.in/2015/0809_pd/secc-excludes-more-it-includes-amidst-high-rural-distress.
10. Jim Yardley, First Hazily, India’s Youth See a Political Role, N.Y. Times (Sept. 1, 2011), https://www.nytimes.com/2011/09/02/world/asia/02india.html.
11. World Bank, South Asia’s Digital Economy: An Opportunity to Build Back Better, Digitally (2022), https://documents1.worldbank.org/curated/en/099230004062228270/pdf/P1723000e5e0d20908c790a5ffdda147f1.pdf.
12. OECD (2024), Modernising Access to Social Protection: Strategies, Technologies and Data Advances in OECD Countries, OECD Publishing, Paris, https://doi.org/10.1787/af31746d-en .
13. U.N. Rsch. Inst. for Soc. Dev., Combating Poverty and Inequality: Structural Change, Social Policy and Politics, U.N. Sales No. E.10.III.Y.1 (2010).
14. Constitutional Law of India – Chapter 7 – Right to Equality, Student Manupatra, http://student.manupatra.com/Academic/Abk/Constitutional-Law-of-India/CHAPTER-7.htm (last visited Feb. 11, 2026).
15. INDIA CONST. art. 15, available at https://www.constitutionofindia.net/articles/article-15-prohibition-of-discrimination-on-grounds-of-religion-race-caste-sex-or-place-of-birth/ (last visited Feb. 11, 2026)
16. Rishika Sahgal, Olga Tellis v. Bombay Municipal Corporation: The Supreme Court of India, the Right to Housing and the Interconnectedness of Rights (June 17, 2022) (unpublished manuscript) (available at University of Oxford).
17. Constitution of India, Part IV: Directive Principles of State Policy, arts. 36–51 (Ministry of External Affairs ed., PDF), available at https://www.mea.gov.in/images/pdf1/part4.pdf (accessed Feb. 11, 2026)
18. People’s Union for Civil Liberties v. Union of India, (2001) 7 S.C.C. 484 (India).
19. Justice K.S. Puttaswamy (Retd.) v. Union of India, (2019) 1 SCC 1 (India) (affirming constitutionality of the Aadhaar (Targeted Delivery of Financial and Other Subsidies, Benefits and Services) Act, 2016)
20. Raj Kumar Reddy, Security Issues in Data Analytics for Financial Services in Security Issues in Data Analytics for Financial Services 10–16 (Vedashree Publishers 2025) (ISBN: 978-81-977245-5-8)
21. Kiruba Sankar Varadharajan, Tinku Thomas & Anura Kurpad, The Indian National Food Security Act, 2013: A Commentary, 35 Food & Nutr. Bull. 253, 253–65 (2014), https://doi.org/10.1177/156482651403500212
22. Lauren Birchfield & Jessica Corsi, The Right to Life Is the Right to Food: People’s Union for Civil Liberties v. Union of India & Others, 17 Hum. Rts. Brief 15 (2010).
23. K.S. Puttaswamy (Retd.) & Anr. v. Union of India & Ors., (2017) 10 SCC 1, (India)
24. Justice K.S. Puttaswamy (Retd.) & Anr. v. Union of India & Ors., (2019) 1 SCC 1 (India) (Supreme Court of India Sept. 26, 2018).
25. Swaraj Abhiyan v. Union of India & Ors., Writ Petition (Civil) No. 857 of 2015, (2016) 7 SCC 498 (India).
26. Priya Rai, R.K. Sharma, P.K. Jain & Akash Singh, Transforming Dimension of IPR: Challenges for New Age Libraries (NLU Delhi Press 2015) (foreword by Ranbir Singh).
27. Protective Discrimination and Social Justice: Article 15 and 16, EduRev (last updated Feb. 7, 2026), https://edurev.in/t/373167/Protective-Discrimination-and-Social-Justice-Article-15-and-16 (last visited Feb. 11, 2026).
28. Yiyang Mei & Michael J. Broyde, Reclaiming Constitutional Authority of Algorithmic Power, arXiv (Aug. 2025), https://doi.org/10.48550/arXiv.2508.11699 .
29. U. Sudhakar Reddy, Telangana Govt’s ‘Samagra Vedika’ Poses Risk to Human Rights, Says Amnesty International, Times of India (June 12, 2024), https://timesofindia.indiatimes.com/city/hyderabad/telangana-governments-samagra-vedika-poses-risk-to-human-rights-says-amnesty-international/articleshow/110923244.cms
30. Oluwabusayo Adijat Bello, Adebola Folorunso, Oluomachi Eunice Ejiofor & Folake Zainab Budale, Machine Learning Approaches for Enhancing Fraud Prevention in Financial Transactions, 10 Int’l J. Mgmt. Tech. 85 (2023), https://doi.org/10.37745/ijmt.2013/vol10n185109
31. Election Commission uploads on website names of people on logical discrepancies list following Supreme Court order, The Hindu (Jan. 24, 2026), https://www.thehindu.com/news/national/west-bengal/election-commission-uploads-on-website-names-of-people-on-logical-discrepancies-list-following-supreme-court-order/article70547602.ece.
32. Faulty Algorithm Denies Food to Thousands of Poor in Telangana, The Reporters’ Collective (Jan. 24, 2024), https://www.reporters-collective.in/newsletters/ai-part-1-newsletter
33. Ministry of Rural Development, Government of India, MGNREGA Sameeksha: An Anthology of Research Studies on the Mahatma Gandhi National Rural Employment Guarantee Act, 2005, 2006–2012, edited and compiled by Mihir Shah, Neelakshi Mann & Varad Pande (New Delhi: Orient BlackSwan, 2012).
34. Delhi Police HCM Typing Test Passage April Month-RKC, Scribd (uploaded by eepwddelhinwr1), 45K views, 404 pages, https://www.scribd.com/document/748065032/Delhi-Police-HCM-Typing-Test-Passage-April-Month-RKC-1 (last visited Feb. 11, 2026)
35. Shiv Sahay Singh, Whiff of Starvation in Jharkhand Deaths, The Hindu (Dec. 3, 2021), https://www.thehindu.com/news/national/whiff-of-starvation-in-jharkhand-deaths/article62031182.ece
36. Vision IAS, Government Schemes Comprehensive Part 2 (May 2025) (PDF), available at https://cdn.visionias.in/value_added_material/e7b24-government-schemes-comprehensive-part-2_2025.pdf (last accessed Feb. 11, 2026)
37. Ayushmaan, Aadhar and Algorithmic Exclusion from Welfare: Case Study from Jharkhand (Dec. 31, 2025), https://nickledanddimed.com/2025/12/31/aadhar-and-algorithmic-exclusion-from-welfare-case-study-from-jharkhand/
38. Sonia Katyal, Private Accountability in the Age of Artificial Intelligence (unpublished manuscript), arXiv:2601.17013 (Jan. 14, 2026) (arguing that opaque private AI systems driven by proprietary technology hinder transparency and accountability and necessitate new forms of oversight to address algorithmic bias and accountability challenges). arXiv preprint available at https://arxiv.org/abs/2601.17013. arXiv:2601.17013 (2026)
39. Aadarsh Anand, Algorithmic Discrimination in India’s Legal System: Constitutional Challenges and Policy Reform, Virtuosity Legal (Oct. 13, 2025), https://virtuositylegal.com/algorithmic-discrimination-in-indias-legal-system-constitutional-challenges-and-policy-reform/.
40. Amal Singh Patel, Algorithmic Bias and the Quest for Equal Justice in India, Virtuosity Legal (Aug. 2025), https://www.virtuositylegal.com/august-2025/algorithmic-bias-india
41. Paul Egan, Michigan Integrated Data Automated System Experiences 93 Percent Error Rate During Nearly Two Years of Operation, Gov’t Tech (July 31, 2017), https://www.govtech.com/data/michigan-integrated-data-automated-system-experiences-93-percent-error-rate-during-nearly-two-years-of-operation.html
42. Urshita Bansal, Sunita Bharatwal & Dhana Bagiyam & Early Ridho Kismawadi, Fraud Detection in the Era of AI: Harnessing Technology for a Safer Digital Economy, in AI-Driven Decentralized Finance and the Future of Finance 139 (2024) (Avinashilingam University & Institut Agama Islam Negeri Langsa) https://doi.org/10.4018/979-8-3693-6321-8.ch006 .
43. Bryce Goodman & Seth Flaxman, European Union Regulations on Algorithmic Decision-Making and a “Right to Explanation”, 38 AI Mag. 50 (2017), https://doi.org/10.1609/aimag.v38i3.2741
44. NCJM et al. v. State of the Netherlands (SyRI), ECLI:NL:RBDHA:2020:865 (Dist. Ct. The Hague Feb. 5, 2020) (concluding that legislation authorizing the System Risk Indication (“SyRI”) welfare-fraud algorithm failed to satisfy transparency and proportionality requirements under applicable fundamental rights law)
45. Magalhães, B. (2025). Scripted to fail: delayed discrimination in biometric asylum systems. Information, Communication & Society, 1–20. https://doi.org/10.1080/1369118X.2025.2561063 .
46. Julia M. Pescarini et al., Evaluating the Impact of the Bolsa Familia Conditional Cash Transfer Program on Premature Cardiovascular and All-Cause Mortality Using the 100 Million Brazilian Cohort: A Natural Experiment Study Protocol, 10 BMJ OPEN e039658 (2020), https://doi.org/10.1136/bmjopen-2020-039658
47. Human Rights Watch, “No Internet Means No Work, No Pay, No Food”: Internet Shutdowns Deny Access to Basic Rights in “Digital India” (June 14, 2023), https://www.hrw.org/report/2023/06/14/no-internet-means-no-work-no-pay-no-food/internet-shutdowns-deny-access-basic
48. Eileen Guo, Gabriel Geiger & Justin-Casimir Braun, Inside Amsterdam’s High-Stakes Experiment to Create Fair Welfare AI, MIT Tech. Rev. (June 11, 2025), https://www.technologyreview.com/2025/06/11/1118233/amsterdam-fair-welfare-ai-discriminatory-algorithms-failure
49. Adamantia Rachovitsa & Niclas Johann, Human Rights Implications of AI in the Digital Welfare State: Lessons from the Dutch SyRI Case, 22 Hum. Rts. L. Rev. 1, 2-3 (2022); GDPR, art. 22, 2016 O.J. (L 119) 1 (EU); Digital Personal Data Protection Act, 2023, No. 22 of 2023 (Ind.).
50. Sajitha Bashir, Carl J. Dahlman, Naoto Kanehira & Klaus Tilmes, The Converging Technology Revolution and Human Capital: Potential and Implications for South Asia (South Asia Dev. Forum 2021), https://doi.org/10.1596/978-1-4648-1719-9 .
51. Kumar Sambhav Shrivastava, How We Investigated Welfare Algorithms in India (Part I), Pulitzer Ctr. (May 30, 2024), https://pulitzercenter.org/how-we-investigated-welfare-algorithms-india-part-i.
52. Silvia Masiero, Explaining Trust in Large Biometric Infrastructures: A Critical Realist Case Study of India’s Aadhaar Project, 84 Electronic J. of Information Sys. in Dev. Countries e12053 (2018), https://doi.org/10.1002/isd2.12053
53. Comm. on Reforms of the Criminal Justice Sys., Ministry of Home Affairs, Gov’t of India, Report of the Committee on Reforms of the Criminal Justice System (2003).
54. World Bank, Enhancing Government Effectiveness and Transparency: The Fight Against Corruption (Sept. 2020), https://documents1.worldbank.org/curated/en/235541600116631094/pdf/Enhancing-Government-Effectiveness-and-Transparency-The-Fight-Against-Corruption.pdf
55. Tarangini Sriraman, The Document in the Digital, in IN PURSUIT OF PROOF: A HISTORY OF IDENTIFICATION DOCUMENTS IN INDIA 209 (Oxford Univ. Press 2018).
56. Sylvia Luv, Data Privacy, Human Rights, and Algorithmic Opacity (2022), https://doi.org/10.15779/Z38804XM07.
57. Rethink Aadhaar, Digital coercion causes one more hunger death; APAAR ID derails education, (Oct. 30, 2025), https://rethinkaadhaar.in/blog/2025/10/30/digital-coercion-causes-one-more-hunger-deathnbsp-apaar-id-derails-educationnbsp.
58. Umar Bashir Mir, Arpan K. Kar & M.P. Gupta, Analysing the Design and Implementation Goals for Aadhaar – The Digital Identity, India Science, Technology & Innovation Portal (May 1, 2024), https://www.indiascienceandtechnology.gov.in/featured-science/analysing-design-and-implementation-goals-aadhaar-%E2%80%93-digital-identity (last visited Feb. 12, 2026)
59. Andreas Holzem, Chapter 4: Confessional Societies and the Christianization of Living Environments, in Christianity in Germany 1550–1850 349–615 (Charlotte P. Kieslich & Ansgar Hastenpflug transs., Brill 2024) (doi: https://doi.org/10.30965/9783657795239_005
60. Jayati Ghosh, Social Protection Programmes in India: An Overview of Recent Experiences with Different Types of Schemes, Governance & Policy Brief (Food and Agriculture Organization of the United Nations, Apr. 2014) (on file with IPC-Info)
61. Watchdoq, India Aadhaar Glitches Linked to Starvation Deaths: A Digital ID Nightmare Exposed, Watchdoq (Nov. 18, 2025), https://watchdoq.com/healthbeat/india-aadhaar-glitches-linked-to-starvation-deaths.
62. Aniket Doegar & Madhura Karnik, Welfare Schemes Don’t Work if Nobody Knows About Them: Fixing India’s Information Deficit, NextBillion (Nov. 30, 2017), https://nextbillion.net/welfare-schemes-dont-work-if-nobody-knows-about-them-fixing-indias-information-deficit/
63. Amarnath Tewary, Jharkhand, Where Not Having Aadhaar Could Starve You to Death, The Hindu (June 9, 2018), https://www.thehindu.com/news/national/other-states/jharkhand-where-not-having-aadhaar-could-starve-you-to-death/article61827611.ece.
64. KHERA, R. (2017). Impact of Aadhaar on Welfare Programmes. Economic and Political Weekly, 52(50), 61–70. http://www.jstor.org/stable/45132600 .
65. Manmeet Kaur et al., Exploring the Complexities of Slum Vulnerability in Haryana, India: A Qualitative Research Journey Into Economic, Social, Physical, and Health Dimensions, 20 Int’l J. Qual. Stud. Health & Well-Being 1 (2024), https://doi.org/10.1080/17482631.2024.2432692 .
66. OECD, Insolvency Systems in Asia: An Efficiency Perspective (OECD Publ’g 2001), https://doi.org/10.1787/9789264189539-en
67. U.N. Women, Preventing Conflict, Transforming Justice, Securing the Peace: A Global Study on the Implementation of United Nations Security Council Resolution 1325 (2015), http://wps.unwomen.org/en.
68. OECD, Artificial Intelligence, Data and Competition, OECD Artificial Intelligence Papers No. 18, OECD Publishing, Paris (May 2024), https://doi.org/10.1787/e7e88884-en.
69. Shaikha Khalfan Saad, Reducing Corruption in Human Resource Management to Improve Employee Quality and Organizational Growth (May 2025) (Ph.D. thesis, University of Córdoba) (on file with University of Córdoba).
70. Global Subsidies Initiative-IISD, Bangladesh Institute of Development Studies (BIDS), Integrated Research and Action for Development (IRADe) & Spaces for Change, Gender and Fossil Fuel Subsidy Reform: Findings from and Recommendations for Bangladesh, India and Nigeria (2019) (ENERGIA report)
71. Jaijit Bhattacharya ed., Technology in Government (GIFT Publ’g 2006).
72. Madhulika Masih, Sarita Barpanda & Zahra Wynne eds., India’s Family Planning Programme: Cruelty in Sterilization Camps—Mistreatment and Coercion: Unethical Sterilization in India (Socio Legal Info. Ctr. 2012)
73. Born, A. M. (2025). Meritocracy from below: dreams, divisions, and the struggle for merit in a stigmatised neighbourhood. City, 29(3–4), 438–458. https://doi.org/10.1080/13604813.2025.2512625 .