Abstract
The rapid integration of artificial intelligence and algorithmic systems into Indian public administration marks a structural transformation in the exercise of state power. From faceless tax assessments and automated welfare eligibility determinations to predictive policing and biometric authentication under Aadhaar, algorithms increasingly perform functions traditionally reserved for human administrative discretion. This article examines whether such algorithmic governance is compatible with the constitutional guarantees of equality under Article 14 and life and personal liberty under Article 21 of the Constitution of India. Despite the increasing adoption of artificial intelligence in public administration, limited constitutional scholarship has examined whether algorithmic decision-making can legitimately replace administrative discretion while remaining consistent with the guarantees of Articles 14 and 21.
The study adopts a doctrinal and comparative legal research methodology through an analysis of constitutional provisions, judicial precedents, statutory instruments, and comparative regulatory frameworks. Drawing upon landmark jurisprudence including E.P. Royappa, Maneka Gandhi, and Puttaswamy, the article argues that fully automated decision-making that lacks transparency, human oversight, and reasoned justification risks violating the doctrines of non-arbitrariness, natural justice, and procedural fairness. Comparative insights from the European Union’s GDPR and AI Act further illuminate the accountability deficits in India’s current framework. The article concludes that artificial intelligence cannot constitutionally replace administrative discretion; it may only assist administrative decision-making under a robust regime of technological legibility, mandatory human review, and rights-based safeguards
Keywords: Algorithmic Governance, Administrative Discretion, Article 14, Article 21, Artificial Intelligence, Natural Justice, Digital Personal Data Protection Act, Judicial Review
I. Introduction
A. Background
Administrative discretion has long been recognised as an indispensable feature of modern governance. Legislatures routinely confer discretionary powers upon executive authorities to enable flexible and context-sensitive application of statutory norms. Indian administrative law has developed sophisticated doctrines to discipline this discretion, insisting that it be exercised reasonably, fairly, and without arbitrariness.[1] The rise of algorithmic systems, however, fundamentally alters the architecture of this discretion. Decision-making authority is increasingly relocated from accountable human officers to opaque computational processes whose internal logic resists conventional modes of judicial scrutiny.
India’s digital transformation has accelerated this shift. [2]The Faceless Assessment Scheme under the Income Tax Act eliminates human interface in tax scrutiny through automated allocation. [3]Aadhaar-based biometric authentication governs access to welfare benefits, pensions, and subsidies. Predictive tools and facial recognition systems are deployed in policing, while automated risk-scoring influences regulatory and quasi-judicial processes. These developments promise efficiency, consistency, and reduction of corruption. Yet they simultaneously raise profound constitutional questions: Can a machine exercise statutory discretion? Does an algorithmic output satisfy the requirements of natural justice? And can the State’s duty to act fairly be discharged by code?
B. Objective
This article undertakes a constitutional examination of algorithmic governance within the Indian administrative framework, focusing on the growing use of artificial intelligence in public decision-making. It explores whether [4]algorithmic systems can lawfully assume functions traditionally exercised through administrative discretion, or whether constitutional guarantees of fairness, equality, and procedural justice require that such discretion remain subject to meaningful human oversight.
To address this question, the study first analyses the [5]constitutional and jurisprudential foundations governing administrative discretion. It then evaluates the compatibility of algorithmic decision-making with the guarantees embodied in Articles 14 and 21 of the Constitution of India. The discussion further draws upon comparative developments in the European Union to identify emerging regulatory approaches before concluding with key findings and proposing a rights-oriented constitutional framework for the responsible integration of artificial intelligence into public administration.
II. Administrative Discretion and Constitutional Discipline
The necessity of administrative discretion stems from the inherent limitations of legislation. While statutes establish the legal framework governing public administration, they cannot anticipate every factual situation or prescribe a rigid solution for every administrative problem. Consequently, legislatures confer discretionary authority upon administrative agencies and public officials to facilitate the effective implementation of statutory objectives. Such discretion enables authorities to respond to diverse factual circumstances, balance competing public interests, and adapt administrative action to changing social, economic, and technological realities. Far from being an exception, discretionary decision-making has become an indispensable component of the modern administrative State.
However, the conferment of discretionary power does not imply the absence of legal constraints. Administrative discretion is exercised within a constitutional framework that requires every public authority to act lawfully, reasonably, and for the purposes for which the power has been conferred. Decisions must be based on relevant considerations, supported by objective reasoning, and free from arbitrariness, mala fides, or improper motives. Equally important is the obligation to exclude irrelevant considerations and ensure that discretion is exercised consistently rather than according to personal preferences or political expediency. These limitations distinguish lawful discretion from arbitrary power and preserve public confidence in administrative institutions.
The evolution of Indian administrative law reflects a gradual but significant expansion of judicial control over discretionary powers. Initially, judicial review was largely confined to examining whether administrative authorities acted within the scope of their statutory powers. Over time, constitutional jurisprudence broadened the scope of review to include standards of fairness, proportionality, transparency, and procedural propriety. The judiciary increasingly recognised that the legitimacy of administrative action depends not only upon the existence of legal authority but also upon the manner in which that authority is exercised. Consequently, constitutional principles have transformed administrative discretion from a matter of executive convenience into a legally accountable exercise of public power.
These foundational principles assume that discretionary decisions are made by human decision-makers capable of evaluating competing considerations, appreciating the factual context of individual cases, exercising independent judgment, and providing intelligible reasons for their conclusions. The emergence of algorithmic governance fundamentally challenges these assumptions. Unlike human administrators, algorithmic systems operate through computational models and statistical inferences that may not adequately account for contextual nuances or explain the reasoning underlying their outputs. This transition from human judgment to automated decision-making raises important constitutional concerns regarding accountability, transparency, and procedural fairness, making it necessary to reconsider whether traditional doctrines governing administrative discretion remain sufficient in an era increasingly shaped by artificial intelligence.
III. Algorithmic Decision-Making under Articles 14 and 21
A. Article 14 and the Prohibition of Arbitrariness
[6]Article 14 guarantees equality before the law and equal protection of the laws.6 Algorithmic
systems pose distinctive challenges to this guarantee. First, opacity often described as the “black box” problem prevents affected persons and reviewing courts from ascertaining whether the decision rests on relevant statutory criteria or on hidden correlations that may embed historical bias.
Second, algorithmic classification frequently relies on proxies that track protected characteristics such as caste, gender, or socio-economic status, producing disparate impacts that may constitute indirect discrimination. Third, the absence of intelligible differentia in probabilistic models undermines the classic two-pronged test of reasonable classification.
The doctrine of manifest arbitrariness, refined in Shayara Bano, provides an additional avenue of challenge.[7] A decision generated by an invalidated or poorly calibrated algorithm that systematically excludes eligible beneficiaries or targets individuals without rational basis may be characterised as capricious and therefore unconstitutional.
The Faceless Assessment Scheme illustrates the tension: while courts have insisted on strict adherence to automated allocation to prevent human interference, they have yet to demand corresponding transparency regarding the algorithms that select cases for scrutiny.
B. Article 21, Natural Justice and Procedural Fairness
Article 21 protects life and personal liberty and, after Maneka Gandhi, requires that any deprivation occur through a procedure that is fair, just and reasonable.[8] The principles of natural justice audi alteram partem and the rule against bias form part of this procedural guarantee.
Automated systems that deny benefits, initiate enforcement action, or restrict liberty without prior notice, opportunity of hearing, or intelligible reasons strike at the heart of these protections.
The Aadhaar litigation remains the most significant judicial engagement with algorithmic exclusion. In Puttaswamy (Aadhaar), the majority upheld the Aadhaar project subject to safeguards, yet acknowledged the problem of authentication failures that excluded genuine beneficiaries.[9] Subsequent experience with welfare automation has confirmed that technical errors biometric mismatches, database inconsistencies, and algorithmic misclassification translate into denial of constitutional entitlements. When the State places the burden of proving identity or eligibility upon the citizen in the face of systemic technological failure, it reverses the ordinary presumption of fairness and violates the spirit of Article 21.
Moreover, the duty to give reasons cannot be satisfied by a bare algorithmic output. A reasoned order must disclose the decisive factors and the process of reasoning. Machine-learning models that operate through multi-layered neural networks often cannot produce such explanations in a form intelligible to the affected person or to a reviewing court. Without technological legibility, judicial review itself is impaired, rendering the constitutional guarantee of effective remedy illusory.
IV. Comparative Perspectives and the Indian Regulatory Landscape
The European Union has confronted similar challenges through a dual regulatory architecture. Article 22 of the GDPR grants individuals the right not to be subject to a decision based solely on automated processing that produces legal or similarly significant effects, subject to limited exceptions and accompanied by rights to human intervention, expression of views, and contestation.[10] The EU Artificial Intelligence Act further classifies certain AI systems used in public administration, law enforcement, and access to essential services as high-risk, imposing obligations of transparency, human oversight, and conformity assessment.[11] These instruments rest on the premise that fully automated decisions affecting fundamental rights require heightened safeguards.
India lacks an equivalent comprehensive framework. The Digital Personal Data Protection Act, 2023, regulates the processing of personal data and imposes obligations on data fiduciaries, including significant data fiduciaries, yet it does not expressly prohibit solely automated decision- making or create a right to explanation. [12] Governmental instrumentalities enjoy broad exemptions. Sectoral initiatives, such as the Supreme Court’s draft Regulations for Use of Artificial Intelligence in Courts, 2026, adopt a cautious approach by prohibiting risk-scoring and insisting that AI remain strictly assistive.13 These developments signal judicial awareness of the risks, but they do not yet constitute a coherent constitutional doctrine applicable across the administrative state.
V. Findings and Discussion
Several conclusions emerge from the foregoing analysis. First, algorithmic systems that produce binding administrative outcomes without meaningful human oversight cannot satisfy the constitutional requirement of non-arbitrariness under Article 14. The absence of intelligible criteria and the presence of potential bias render such decisions vulnerable to challenge. Second, the principles of natural justice and the duty to give reasons under Article 21 demand that affected persons receive notice, an opportunity to be heard, and an explanation capable of being understood and contested. Purely algorithmic outputs fail this test. Third, the traditional Wednesbury standard of unreasonableness is ill-suited to the opacity and complexity of machine-learning systems; a more structured proportionality-oriented review incorporating explainability and auditability is required[13].
[14]Fourth, the current statutory landscape, centred on the Digital Personal Data Protection Act, 2023 and fragmented sectoral guidelines, leaves a significant accountability vacuum. Without express legislative authorisation and accompanying safeguards, the delegation of statutory discretion to automated systems risks being ultra vires. Finally, comparative experience demonstrates that it is possible to harness the efficiency gains of AI while preserving human responsibility and individual rights. India’s constitutional culture, with its emphasis on substantive due process and anti-arbitrariness, provides fertile ground for developing an indigenous doctrine of technological legibility.
VI. Conclusion
Artificial intelligence cannot constitutionally replace administrative discretion in India. Discretion is not merely a technical function of applying rules to facts; it is a constitutionally disciplined exercise of public power that presupposes human judgment, accountability, and the capacity to respond to the particularities of individual cases. Algorithms may assist, inform, and even recommend, but the final responsibility for decisions that affect life, liberty, equality, and dignity must remain with human officers who can be held to account under Articles 14 and 21. The challenge before Indian law is not to halt technological progress but to constitutionalise it— to ensure that the efficiency of the machine never eclipses the fairness, reasonableness, and humanity that the Constitution demands of the State. Only through a doctrine of technological legibility, anchored in transparency, human oversight, and robust remedies, can algorithmic governance be reconciled with the rule of law.
REFERENCE
[1] E.P. Royappa v. State of Tamil Nadu, (1974) 4 SCC 3.
[2] Faceless Assessment Scheme, Income Tax Act, 1961.
[3] Justice K.S. Puttaswamy (Retd.) v. Union of India (Aadhaar), (2019) 1 SCC 1
[4] H.W.R. Wade & C.F. Forsyth, Administrative Law, 12th ed. (Oxford University Press, 2024).
[5] M.P. Jain & S.N. Jain, Principles of Administrative Law, 9th ed. (LexisNexis, 2022).
[6] E.P. Royappa v. State of Tamil Nadu, (1974) 4 SCC 3.
[7] Shayara Bano v. Union of India, (2017) 9 SCC 1.
[8] Maneka Gandhi v. Union of India, (1978) 1 SCC 248.
[9] Justice K.S. Puttaswamy (Retd.) v. Union of India (Aadhaar), (2019) 1 SCC 1.
[10] Regulation (EU) 2016/679 (General Data Protection Regulation), art. 22.
[11] Regulation (EU) 2024/1689 (Artificial Intelligence Act).
[12] Supreme Court of India, Draft Regulations for Use of Artificial Intelligence in Courts, 2026.
[13]Supreme Court of India, Draft Regulations for Use of Artificial Intelligence in Courts, 2026.
[14] Constitution of India, art. 21.