Code as State Action: Interrogating Algorithmic Arbitrariness, Indirect Discrimination, and Equality under Articles 14 and 15

Public authorities in India are steadily handing eligibility screening, risk scoring and benefit allocation to software. This paper asks whether the code performing these functions can be treated as state action and, if so, how Articles 14 and 15 of the Constitution constrain it. It argues that the relevant act is not the programmer’s intention but the authority’s decision to adopt, configure and rely on a system, and that Article 12, read with the instrumentality and public-function tests, reaches that decision even where a private vendor builds the model. The paper then develops a theory of algorithmic arbitrariness. Drawing on E.P. Royappa, Maneka Gandhi and Shayara Bano, it contends that an automated determination which is unexplained, unvalidated or unreviewable offends Article 14 because it cannot be measured against any intelligible standard. Turning to discrimination, it shows how facially neutral models reproduce disadvantage through proxy variables and skewed training data, and it uses Lt. Col. Nitisha v. Union of India, India’s leading recognition of indirect discrimination, as the doctrinal bridge to effects-based review under Articles 14 and 15. Reading these authorities alongside Huq’s critique of intent-centred doctrine, the Micklitz volume and Sulmicelli’s safeguards architecture, the paper assesses NITI Aayog’s Responsible AI papers and the 2025 India AI Governance Guidelines and finds them principled but non-justiciable. It closes with four proposals: a presumption of arbitrariness where reasons and validation records are withheld, mandatory equality impact assessments, a right to human review, and aggregate remedies for systemic design defects.