Disclosure of algorithmic decision criteria

Introduction

Disclosure of algorithmic decision criteria refers to the legal obligation or expectation that organisations explain the principal factors, rules, parameters and processes used by automated systems to make decisions affecting individuals or businesses. Such systems are increasingly used in banking, recruitment, healthcare, insurance, education, taxation, policing and digital platforms. They may determine loan eligibility, employment shortlisting, insurance premiums, fraud detection or access to public services.

The central purpose of disclosure is to ensure transparency, fairness, accountability and protection against arbitrary decision-making. Individuals should be able to understand why an algorithm has produced a particular outcome and whether the decision can be challenged. Although India does not yet have one comprehensive law specifically governing disclosure of all algorithmic decision criteria, several constitutional, statutory and judicial principles provide a foundation for such accountability.

Meaning and scope

Algorithmic decision criteria include the variables, datasets, weightings, thresholds, assumptions and decision rules used by an automated system. Disclosure does not necessarily require organisations to publish their entire source code or confidential technical architecture. Instead, it generally concerns meaningful information about the factors that materially influenced a decision.

For example, if a bank rejects a loan application through an automated system, meaningful disclosure may include the relevant eligibility requirements, significant factors affecting the result and the procedure for requesting human review.

The objective is to make decisions understandable without unnecessarily exposing trade secrets, cybersecurity-sensitive information or personal data belonging to others.

Legal framework in India

1. Constitutional protection

Articles 14 and 21 of the Constitution provide important foundations for algorithmic transparency. Article 14 protects against arbitrary state action and requires fairness and non-arbitrariness in public decision-making. Article 21 protects life and personal liberty, including interests connected with privacy, dignity and procedural fairness.

Where government authorities rely on automated systems to distribute benefits, identify beneficiaries or make administrative decisions, unexplained algorithmic outcomes may raise constitutional concerns. Public authorities should be able to demonstrate that their systems operate according to lawful, reasonable and non-discriminatory criteria.

2. Digital Personal Data Protection Act, 2023

The Digital Personal Data Protection Act, 2023 establishes obligations concerning the processing of digital personal data. Its relevance arises when algorithmic systems use personal data to generate decisions. However, it should not be treated as a comprehensive right to obtain algorithmic explanations or source code. Its application also depends on the commencement of the relevant provisions and rules.

3. Information Technology Act, 2000

The Information Technology Act, 2000 provides a broader legal framework for electronic records, digital transactions and cybersecurity. Although it does not establish a general algorithmic disclosure right, its provisions may become relevant where automated systems involve unlawful data practices, unauthorised access or other legally recognised violations.

4. Right to Information Act, 2005

The RTI Act can support transparency where public authorities use algorithms in administrative decision-making. Citizens may seek information about government policies, selection criteria, decision procedures and records concerning automated systems. However, disclosure may be restricted by statutory exemptions, including protections for commercial confidence, privacy and security.

Important case laws

1. Justice K.S. Puttaswamy v. Union of India (2017) 10 SCC 1

The Supreme Court recognised privacy as a fundamental right under Article 21 and other constitutional freedoms. The judgment emphasised informational privacy and individual autonomy. Its principles are relevant where algorithms collect and analyse personal information to make decisions. It supports the need for lawful, proportionate and accountable data processing, although it did not establish a specific right to algorithmic explanations.

2. Maneka Gandhi v. Union of India (1978) 1 SCC 248

The Supreme Court held that a procedure affecting personal liberty must be fair, just and reasonable rather than arbitrary. This principle is significant when public authorities use automated systems to make decisions affecting individual rights. An automated process should not become a mechanism for avoiding procedural fairness or meaningful review.

3. E.P. Royappa v. State of Tamil Nadu (1974) 4 SCC 3

The Supreme Court explained that arbitrariness is incompatible with equality under Article 14. This principle can be applied where algorithmic systems produce inconsistent, discriminatory or unreasonable outcomes in public administration. Authorities must ensure that automated decisions are supported by rational and lawful criteria.

4. Kranti Associates Pvt. Ltd. v. Masood Ahmed Khan (2010) 9 SCC 496

The Supreme Court emphasised the importance of recording reasons in judicial and administrative decisions. Reasoned decisions promote transparency and facilitate review. This principle is particularly relevant when an algorithm assists a public authority in reaching a decision: the authority should be able to explain the basis of its final determination.

Challenges and safeguards

Algorithmic disclosure faces several practical difficulties. Complex machine-learning models may be difficult to interpret, while excessive disclosure may expose confidential business information or create opportunities for manipulation. In addition, biased training data may produce discriminatory outcomes even when the system appears technically neutral.

Effective safeguards should include:

Providing understandable explanations of significant decision factors.

Maintaining records of algorithmic decisions and subsequent human interventions.

Conducting regular bias and accuracy assessments.

Providing accessible complaint and appeal mechanisms.

Ensuring human review for decisions with serious legal or personal consequences.

Protecting trade secrets without concealing information necessary to challenge unfair decisions.

Conclusion

Disclosure of algorithmic decision criteria is an important element of transparent and accountable digital governance. Although Indian law does not currently provide a universal, standalone right to inspect every algorithm or obtain a detailed explanation of every automated outcome, constitutional principles, privacy protections, information-access laws and requirements of reasoned administrative decision-making provide a foundation for accountability. Future regulation should establish clear standards for meaningful explanations, independent audits, human review and effective remedies against discriminatory or arbitrary algorithmic decisions.