Written by Shiny B,
Tamil Nadu National Law University, Thiruchirappalli,
July 2026
Introduction
All constitutions begin by defining the location of power. None foresaw its destination. Athens had its Agora, Rome its Senate, Westminster its Parliament. Constitutional democracies-built courts. The twenty-first century builds data centres. Power has moved, but the law still addresses its official correspondence to the wrong place.
This forwarding error has been particularly pronounced in India’s digital welfare state, where a machine’s decision dictates whether a family eats. This article addresses a different problem from the infrastructure critiques that dominate recent commentary. The issue is not simply that automated welfare systems exclude people, but that they are breaking down the household — the building block of society and statute through which food security, livelihoods and welfare are realised under the National Food Security Act 2013 (NFSA)[1]. For as rights vest formally with the individual but are allocated and delivered via households, the question is not whether machines err, but whether administrative classifications mediated by automation are corrupting the statutory idea of the household and undermining constitutional enjoyment.
It may be useful to move beyond the label “algorithmic governance” and parse discrete layers of data-driven welfare administration: digital identification (creating unique identity profiles), digital authentication (identity verification at the point of transaction)[2], digital eligibility determination (rule-based filtering of beneficiaries through exclusion engines and cross-state databases), and digital enforcement (delivery or refusal of benefit)[3]. Each of these stages imposes friction, but the point at which automated eligibility screens meet real-time verification derails the household unit.
Constitutional Injury: Re-reading Articles 14 and 21
Under Article 14, state action cannot be arbitrary[4], an argument laid down by the Court in E.P. Royappa v. State of Tamil Nadu, where the bench described equality as a “dynamic concept with many aspects and dimensions” which “cannot be ‘cribbed, cabined and confined’, within traditional and doctrinaire limits”. Arbitrariness is at odds with equality[5].
This algorithmic interference is manifested where automated systems substitute proxies for actual metrics of constitutionally relevant need. Data centric models eschew actual material deprivation in favour of proxies held by different government departments[6]. Although proxies are never formally binding evidence of wealth, they have become de facto proxies of eligibility and have been documented as imposing real exclusion costs on the genuinely deserving in numerous studies[7]. In Telangana, a widow’s household was struck off the food security rolls because an eligibility engine wrongly matched her late husband, a rickshaw puller, to a car owner’s record. Such reliance produces data driven arbitrariness: it treats non-conclusive administrative records as absolute truth and denies the citizen any chance to rebut the inference, which under the Royappa standard is itself arbitrary and antithetical to equality.
Data driven systems also violate the terms of Article 21[8], the core constitutional guarantee to the conditions of life, many of which are realised at the household level (subsidised foodgrains are physically and economically consumed collectively). The harms of algorithmic exclusion thus fall indivisibly across a household’s members, a vulnerability reported across multiple field studies where failed biometric authentication[9] led to exclusion from rations; often resulting in hunger, debt at exploitative interest rates from informal creditors, and cycles of intergenerational debt[10], as recorded from ground reporting out of Jharkhand[11].
Statutory Distortion and the Undermining of Gendered Agency
These algorithmic partitions respect neither constitutional law nor the text of the statute itself. The NFSA frames the household as the primary beneficiary; Section 13 explicitly provides that “the eldest woman in the household” be the “head of the household” for purposes of a ration card[12].
This provision implements the protective discrimination mandate of Article 15(3)[13] and was intended to facilitate women’s bargaining power within households. When household profiles are broken, merged, deleted or altered due to simple, clerical mismatches[14]; deletions which appear to occur at a vast scale without ever having engaged the actual statutory structure[15] that gives a family its official head of household status, there is nothing necessarily to indicate that the underlying classification made by the machine bearing no relation to actual deprivation will result in improved outcomes.
The Administrative Vacuum and the Core Concern
The fundamental crisis arising here centres on the law of automated exclusion. Under Maneka Gandhi v. Union of India, any process that deprives someone of their life or liberty must be fair and reasonable; that is, non-arbitrary, non-fanciful, non-oppressive[16]. This requires that due process rights including natural justice and the duty to give reasons must apply.
Automated systems are inherently deficient in meeting this basic due process requirement. An automated system may reject a family’s eligibility, but no clear, coherent explanation reaches the beneficiary; the field staff, unable to change the output, only see a system error such as “Authentication Failed”[17]. Without this rationale, state action becomes opaque, with public officials explaining decisions through the internal logic of the computer, leaving families with no meaningful recourse to appeal. This obscurity mirrors the conditions that led the Supreme Court to insist in Anuradha Bhasin v. Union of India that state action impacting fundamental rights must be public, transparent, and amenable to judicial scrutiny[18]. Notably in Justice K.S. Puttaswamy v. Union of India, the Court sanctioned Aadhaar based disbursement conditional on the direction that ‘no genuine beneficiary shall be denied benefits for want of authentication’[19], a judicial condition of substantive fairness which is now often overridden by strict algorithmic enforcement[20].
Institutional Remedies
To secure constitutional and administrative sovereignty over automated administrative decision-making, the state must dismantle the myth of computational omnipotence and rebuild an infrastructure for humans in the loop systems of decision making. This means incorporating:
Mandatory human intervention. Automated decision-making engines should not be able to automatically remove a household profile from the welfare system without prior human approval.
Intelligible administrative outcomes. Automated systems must be mandated by law to provide written explanations for adverse decisions, indicating what was the exact computer driven process that led to exclusion.
Localized redressal. Well-staffed grievance forums, envisaged by the NFSA[21], should provide immediate human review of erroneously rejected applications.
Right to a human decision. No decision affecting basic survival should rest solely on automated processing; a right to meaningful human review of adverse determinations should be read into Article 21 as a condition of the right to life and livelihood.
Conclusion
At heart, the problem is not the automation of administration, but the fact that automated decision-making is being treated as a standalone basis for substantive rights, without the processual guarantees that have traditionally accompanied administrative actions that directly affect the citizen. Constitutionalism demands that the state use automation as a tool of administration rather than allow it to perform the job of determination of rights on its own authority, lest the constitutional values of equality and dignity become what no one at the time of crafting imagined: a successful database query.
References
[1] The National Food Security Act, 2013, No. 20, Acts of Parliament, 2013 (India).
[2] The Aadhaar (Targeted Delivery of Financial and Other Subsidies, Benefits and Services) Act, 2016, No. 18, Acts of Parliament, 2016, § 7 (India).
[3] See Dalberg, State of Aadhaar 2019: A People’s Perspective (2019).
[4] India Const. art. 14.
[5] E.P. Royappa v. State of Tamil Nadu, (1974) 4 SCC 3 (India).
[6] See Tapasya, Kumar Sambhav & Divij Joshi, How an Algorithm Denied Food to Thousands of Poor in India’s Telangana, Al Jazeera (Jan. 24, 2024).
[7] Karthik Muralidharan, Paul Niehaus & Sandip Sukhtankar, Identity Verification Standards in Welfare Programs: Experimental Evidence from India, 107(2) Rev. Econ. & Stat. 372 (2025); see also O. Grace Ngullie, Examining Exclusions in the Public Distribution System, 6(1) Indian Pub. Pol’y Rev. 123 (2025).
[8] India Const. art. 21.
[9] Jean Drèze et al., Aadhaar and Food Security in Jharkhand: Pain Without Gain?, 52(50) Econ. & Pol. Wkly. 50 (2017).
[10] See LibTech India, Length of the Last Mile: Delays and Hurdles in NREGA Wage Payments (2020).
[11] Nileena M.S., How Aadhaar and Digitisation Compounded Problems Plaguing the PDS in Jharkhand, The Caravan (Aug. 1, 2017).
[12] NFSA, supra note 1, § 13(1).
[13] India Const. art. 15, cl. 3.
[14] See LibTech India, MGNREGA Implementation in India: Insights and Trends, April–September 2024 (2024).
[15] LibTech India, The Missing Work: A National Review of MGNREGA Implementation, FY 2024–25 (2025).
[16] Maneka Gandhi v. Union of India, (1978) 1 SCC 248 (India).
[17] See LibTech India, MGNREGA as a Technological Laboratory (Working Paper, 2023).
[18] Anuradha Bhasin v. Union of India, (2020) 3 SCC 637 (India).
[19] Justice K.S. Puttaswamy (Retd.) v. Union of India, (2019) 1 SCC 1 (India).
[20] See Drèze et al., supra note 9; Nileena M.S., supra note 11.
[21] NFSA, supra note 1, §§ 14–15.