ALGORITHMIC BIAS AND ARTICLE 14

This Article is written by Falak Imam, of Jamia Hamdard University, an intern under Legal Vidhiya.

ABSTRACT

The rapid integration of algorithm-driven decision-making into governmental and private spheres has generated serious constitutional concerns relating to equality, fairness, and non-arbitrariness. Algorithms now influence welfare allocation, policing, surveillance, employment, and credit assessment. While often presented as neutral and efficient, these systems frequently reproduce and intensify existing social inequalities due to biased data, opaque design, and discriminatory outcomes. Within the Indian constitutional framework, Article 14 guarantees equality before the law and equal protection of the laws and operates as a safeguard against arbitrary state action. This article examines whether algorithmic bias can amount to a violation of Article 14 and assesses the applicability of constitutional principles to automated governance. By analysing Indian constitutional jurisprudence, key Supreme Court decisions, and comparative perspectives, the article argues that unregulated algorithmic governance poses a serious threat to substantive equality and demands urgent constitutional scrutiny.

Keywords

Algorithmic Bias, Article 14, Equality, Artificial Intelligence, Non-Arbitrariness, Indian Constitutional Law, Automated Decision-Making, Discrimination

Introduction

The rapid digitisation of governance and administrative decision-making has fundamentally altered the manner in which power is exercised by the State. Across the globe, governments increasingly rely on algorithmic and artificial intelligence–driven systems to manage large populations, allocate public resources, and maintain law and order. In India, this shift is particularly visible in areas such as welfare distribution, biometric identification, predictive policing, surveillance, and eligibility determination for public benefits. Parallel to this, private actors have also embraced automated decision-making systems for recruitment, credit scoring, insurance assessment, and content moderation. While these technologies are often justified on the grounds of efficiency, speed, and objectivity, their growing influence raises serious constitutional concerns.

In a society marked by deep-rooted structural inequalities based on caste, gender, religion, disability, and socio-economic status, the use of algorithmic systems carries the risk of reproducing and intensifying existing forms of discrimination. The assumption that algorithms are neutral or value-free is increasingly being questioned, as empirical evidence reveals that automated systems frequently reflect the biases present in the data on which they are trained and the choices made by their designers. When such systems are adopted or relied upon by the State, the resulting outcomes directly implicate constitutional guarantees of equality and fairness.

Article 14 of the Constitution of India, which guarantees equality before the law and equal protection of the laws, occupies a central position in India’s constitutional framework. Judicial interpretation has consistently expanded the scope of Article 14, transforming it from a formal guarantee of equality into a substantive safeguard against arbitrariness, unreasonableness, and discrimination. This article examines whether algorithmic bias, particularly when embedded in state decision-making, can amount to a violation of Article 14. It argues that unregulated algorithmic governance poses a serious threat to substantive equality and must therefore be subjected to rigorous constitutional scrutiny.

Understanding Algorithmic Bias

Algorithmic bias refers to persistent and unjust patterns of discrimination produced by automated decision-making systems, resulting in unequal treatment of individuals or groups. Contrary to popular belief, algorithms do not operate independently of human influence. They are created, trained, and deployed by humans, and are therefore shaped by social, cultural, and institutional contexts. As a result, algorithmic systems often mirror the prejudices and power structures that already exist within society.

Bias can enter algorithmic systems at multiple stages. Historical bias arises when past data reflects discriminatory practices or unequal social conditions, which are then encoded into datasets used for training algorithms. Design bias emerges from subjective choices made by developers regarding model architecture, variable selection, and optimisation goals. Deployment bias occurs when algorithms are used in contexts different from those for which they were originally designed, often without adequate adaptation. Over time, feedback loops may further entrench bias by reinforcing discriminatory outcomes through repeated use. For example, an employment algorithm trained on historical hiring data that favoured male candidates may systematically disadvantage women, thereby perpetuating gender inequality under the guise of technological efficiency.

Algorithmic bias can manifest in several distinct forms. Historical bias reflects pre-existing social inequalities embedded in data. Representation bias arises when certain communities are underrepresented or misrepresented in datasets. Measurement bias results from the use of flawed or oversimplified indicators to assess complex human attributes. Automation bias refers to the tendency of human decision-makers to place undue trust in algorithmic outputs, often without critical evaluation. When such biases influence decisions relating to access to rights, benefits, or essential public services, they raise serious constitutional concerns.

Algorithmic Governance in India

India has witnessed a rapid expansion of algorithmic governance across multiple sectors. Aadhaar-based biometric authentication has become central to the delivery of welfare schemes, while predictive policing tools and facial recognition technologies are increasingly used for law enforcement and surveillance. Automated systems are also employed to determine eligibility for subsidies, scholarships, and other public benefits. These technologies are frequently promoted as tools to enhance efficiency, reduce corruption, and improve administrative accuracy.

However, the deployment of such systems has also led to significant exclusionary outcomes. Biometric authentication failures under Aadhaar have resulted in the denial of food rations and social security benefits, disproportionately affecting elderly individuals, persons with disabilities, and marginalised communities. Predictive policing tools risk reinforcing existing patterns of over-policing in certain neighbourhoods, while facial recognition technologies have been criticised for higher error rates in identifying minorities. These outcomes highlight the dangers of relying on algorithmic systems without adequate safeguards of transparency, accountability, and human oversight.

Article 14: Constitutional Framework

Article 14 of the Constitution of India provides that the State shall not deny to any person equality before the law or equal protection of the laws within the territory of India. This provision embodies both a negative injunction against arbitrary state action and a positive obligation to ensure equal treatment of similarly situated persons. Over time, the Supreme Court has expanded the interpretation of Article 14 to include principles of fairness, reasonableness, and substantive equality.

Traditionally, Article 14 permitted classification provided that such classification was based on an intelligible differentia and bore a rational nexus with the object sought to be achieved. However, this formal approach was significantly broadened in E.P. Royappa v. State of Tamil Nadu, where the Supreme Court held that arbitrariness is antithetical to equality. This shift marked a departure from rigid classification-based analysis and opened the door to a more substantive understanding of equality. The decision in Maneka Gandhi v. Union of India further clarified that even in the absence of classification, arbitrary and unreasonable state action would violate Article 14.

These developments are particularly relevant in the context of algorithmic governance. When the State relies on automated systems to make decisions that affect individuals’ rights and entitlements, the constitutional focus must shift from the form of decision-making to its impact. If algorithmic systems produce discriminatory or arbitrary outcomes, they cannot be insulated from constitutional scrutiny merely because they are technologically mediated.

ALGORITHMIC BIAS AS ARBITRARINESS AND THE NEED FOR CONSTITUTIONAL RESPONSE

Algorithmic Bias as Arbitrary State Action

Algorithmic decision-making may infringe Article 14 in several ways. Where decision-making processes lack transparency, affected individuals are denied access to reasons or explanations for adverse outcomes. When individuals are deprived of an opportunity to challenge or appeal algorithmic decisions, procedural fairness is undermined. Moreover, where algorithmic outcomes disproportionately disadvantage specific groups, the principle of substantive equality is violated.

When the State adopts or relies upon algorithmic systems that generate discriminatory outcomes without adequate safeguards, such action may amount to manifest arbitrariness, as recognised by the Supreme Court in Shayara Bano v. Union of India. The opacity of many algorithmic systems, often described as “black boxes,” makes it difficult for individuals to understand how decisions affecting them are made. This lack of transparency directly conflicts with principles of natural justice and procedural fairness that are integral to Article 14.

In Kranti Associates v. Masood Ahmed Khan, the Supreme Court emphasised that reasoned decision-making is an essential component of fairness and accountability in administrative action. Algorithmic systems that fail to provide intelligible explanations for their decisions undermine this requirement and weaken the possibility of meaningful judicial review.

Disparate Impact and Substantive Equality

Even when algorithms appear neutral on their face, they may produce outcomes that disproportionately harm marginalised communities. Contemporary constitutional interpretation increasingly recognises indirect discrimination, where neutral policies have unequal effects. In Anuj Garg v. Hotel Association of India, the Supreme Court rejected laws based on stereotypical assumptions and emphasised the need for substantive equality. Algorithmic systems built upon biased data or flawed assumptions risk reinforcing systemic disadvantage and perpetuating inequality.

The Supreme Court’s recognition of indirect and systemic discrimination in cases such as Navtej Singh Johar v. Union of India further strengthens the argument that algorithmic bias falls within the ambit of Article 14. Predictive policing algorithms that disproportionately target certain neighbourhoods, credit-scoring systems that disadvantage economically weaker sections, and facial recognition technologies that misidentify minorities all raise serious concerns regarding substantive equality.

Judicial Review of Algorithmic Systems

When the State deploys algorithms, it effectively delegates decision-making authority to technological systems. Such delegation raises constitutional concerns, particularly when it occurs without adequate safeguards. In Ajoy Kumar Banerjee v. Union of India, the Supreme Court held that excessive delegation without clear policy guidance is unconstitutional. Applying this principle, algorithmic systems that operate without transparency, accountability, or clear standards may amount to unconstitutional delegation.

A core component of Article 14 is the requirement that state action must be reasonable, fair, and non-arbitrary. This includes the duty to provide reasons. Algorithmic opacity undermines the right to challenge adverse decisions, weakens procedural fairness, and restricts the scope of meaningful judicial review. As a result, unexplainable algorithmic decision-making is fundamentally incompatible with Article 14.

Limitations of Article 14 in Addressing Algorithmic Bias

Despite its expansive interpretation, Article 14 faces certain limitations in addressing algorithmic bias. India currently lacks a comprehensive statutory framework regulating artificial intelligence and automated decision-making. Courts are therefore compelled to rely on broad constitutional principles rather than precise legislative safeguards. Additionally, proving algorithmic bias presents significant evidentiary challenges due to proprietary protections, trade secrets, and technical complexity. Judicial capacity constraints further complicate effective scrutiny of complex algorithmic systems.

Comparative Perspectives and the Way Forward

Comparative constitutional developments offer valuable guidance. The European Union’s General Data Protection Regulation and proposed Artificial Intelligence Act adopt a rights-based approach emphasising transparency, accountability, and the right to explanation. These principles resonate strongly with Article 14’s emphasis on fairness and non-arbitrariness. In contrast, the United States relies primarily on statutory anti-discrimination frameworks, which limits the scope of constitutional remedies.

India can draw from these experiences by adopting mandatory algorithmic impact assessments, transparency obligations, and human-in-the-loop safeguards. Judicial guidelines, similar to those laid down in Vishaka v. State of Rajasthan, could provide interim protection in the absence of comprehensive legislation.

Conclusion

Algorithmic decision-making represents a profound shift in governance, offering efficiency while posing serious risks to constitutional equality. In a society characterised by deep structural inequalities, unregulated algorithmic systems threaten to entrench discrimination under the guise of technological neutrality. Article 14, with its emphasis on non-arbitrariness, fairness, and substantive equality, provides a powerful constitutional tool to challenge algorithmic bias. However, constitutional adjudication alone is insufficient. A holistic response combining judicial oversight, legislative reform, and technological accountability is essential to ensure that algorithms serve constitutional values rather than undermine them.

References

1. INDIA CONST. art. 14.
2. General Data Protection Regulation, Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016, 2016 O.J. (L 119) 1.
3. State of W.B. v. Anwar Ali Sarkar, AIR 1952 SC 75 (India).
4. E.P. Royappa v. State of T.N., (1974) 4 S.C.C. 3 (India).
5. Maneka Gandhi v. Union of India, (1978) 1 S.C.C. 248 (India).
6. Ajoy Kumar Banerjee v. Union of India, (1984) 3 S.C.C. 127 (India).
7. Anuj Garg v. Hotel Ass’n of India, (2008) 3 S.C.C. 1 (India).
8. Kranti Assocs. Pvt. Ltd. v. Masood Ahmed Khan, (2010) 9 S.C.C. 496 (India).
9. Shayara Bano v. Union of India, (2017) 9 S.C.C. 1 (India).
10. Navtej Singh Johar v. Union of India, (2018) 10 S.C.C. 1 (India).
11. Vishaka v. State of Rajasthan, (1997) 6 S.C.C. 241 (India).
12. NITI AAYOG, NATIONAL STRATEGY FOR ARTIFICIAL INTELLIGENCE: #AIforALL (2018).
13. Frank Pasquale, The Black Box Society, 47 HARV. C.R.-C.L. L. REV. 1 (2012).
14. Danielle Keats Citron, Technological Due Process, 85 WASH. U. L. REV. 1249 (2008).
15. Reuben Binns, Fairness in Machine Learning, 81 MOD. L. REV. 1 (2018).
16. U.N. High Comm’r for Human Rights, The Right to Privacy in the Digital Age, U.N. Doc. A/HRC/48/31 (2021).

Disclaimer: The materials provided herein are intended solely for informational purposes. Accessing or using the site or materials does not establish an attorney-client relationship. The information presented on this site is not to be construed as legal or professional advice, and it should not be relied upon for such purposes or used as a substitute for advice from a licensed attorney in your state. Additionally, the viewpoint presented by the author is personal.