The Right to Explanation in Algorithmic Decision-Making: A Fundamental Guarantee of Digital Governance

Artificial intelligence increasingly influences decisions about employment, credit, education, access to public services and other matters that shape people’s lives. These systems may process information at a scale that human decision-makers cannot easily match. Yet efficiency does not answer a basic question: what can a person do when an automated system contributes to a decision that affects them, but the reasons for that decision remain unclear?

The right to explanation should be understood as a practical condition of effective accountability. It does not require organisations to disclose every line of source code or reveal commercially sensitive information. It requires them to provide affected individuals with an intelligible account of the role played by an AI system and the principal factors that shaped an outcome, so that the decision can be understood and, where appropriate, challenged.

A decision is difficult to contest if its grounds are invisible

Consider an applicant rejected for a job, a person denied access to a public benefit, or a customer refused a financial service after an automated assessment. If the individual receives only the outcome, they may be unable to identify whether it resulted from inaccurate data, an unsuitable criterion, a technical error or a pattern that disadvantages a particular group. A formal right to complain offers limited protection when the person does not know what they are expected to challenge.

This is not simply a question of better customer communication. It concerns procedural fairness: people affected by consequential decisions should have a meaningful opportunity to understand the basis of those decisions and to seek review. Explanation can expose errors, make discrimination easier to identify and help a reviewer assess whether the decision was lawful and reasonable.

A real example illustrates this problem. In Austria, a mobile telephone operator refused to conclude or extend a contract with a customer because an automated credit assessment indicated that her creditworthiness was insufficient. The assessment had been carried out by Dun & Bradstreet Austria. The customer sought meaningful information about the logic behind the assessment, but the company failed to provide an adequate explanation. The dispute eventually reached the Court of Justice of the European Union in Dun & Bradstreet Austria (Case C-203/22), demonstrating how the absence of intelligible information can prevent an individual from understanding and effectively challenging an automated decision.

Existing law offers important, but bounded, safeguards

The European Union’s General Data Protection Regulation (GDPR) provides one starting point. Article 22 addresses decisions based solely on automated processing that produce legal effects or similarly significantly affect an individual. In specified circumstances, it restricts such decisions and provides safeguards including the right to obtain human intervention, express one’s point of view and contest the decision. These protections are not a universal right to an explanation of every algorithm-assisted decision. Their scope depends on the conditions and exceptions set out in the Regulation.

The interpretation of these safeguards was further clarified by the Court of Justice of the European Union in Dun & Bradstreet Austria (Case C-203/22, 27 February 2025). The Court held that, in cases of automated decision-making, including profiling, within the meaning of Article 22(1) GDPR, Article 15(1)(h) requires controllers to provide meaningful information about the logic involved. This includes an explanation, in a concise, transparent and intelligible form, of the procedure and principles actually applied in using an individual’s personal data to reach a specific result, such as a credit profile. The judgment makes clear that a meaningful explanation must enable the individual to understand the decision and exercise their rights; merely referring to the use of an algorithm is not sufficient.

The EU Artificial Intelligence Act takes a more targeted step. Article 86 gives an affected person, subject to its defined conditions, a right to obtain from the deployer a clear and meaningful explanation of the role of a high-risk AI system in a decision-making procedure and the main elements of the decision. The provision is significant because it connects explanation to particular high-risk uses and consequential decisions. It should not, however, be read as creating an unrestricted right to inspect an AI system or as replacing other rights of appeal and judicial review.

The distinction matters. Legal safeguards should be described accurately: a right to explanation may arise under particular legal frameworks, but its existence, scope and remedies are not identical across jurisdictions.

India presents a different legal position. The Digital Personal Data Protection Act, 2023 (DPDP Act) does not expressly establish a standalone right to obtain an explanation of decisions made through automated processing, nor does it create a general right to human review of such decisions. Although the Act provides for rights concerning access to information about personal data, correction and erasure, and grievance redressal, these protections should not be equated with a specific right to understand the reasoning behind an algorithmic outcome.

Furthermore, under the commencement notification issued by the Government of India on 13 November 2025, the Act’s core provisions, including these individual rights, are scheduled to come into force eighteen months after that date, on 13 May 2027. This highlights the importance of considering not only what a legal framework provides on paper, but also when its protections become operational.

Explanation must lead to meaningful review

A useful explanation should do more than state that an algorithm was involved. Depending on the context, it should identify the system’s role, the principal factors that materially influenced the outcome, relevant data concerns, and the route available to request human reconsideration or challenge the decision. The level of detail should be proportionate to the decision’s impact and the information needed to exercise available rights.

This does not mean that source code must routinely be disclosed. Intellectual property, privacy, security and the prevention of gaming or manipulation may justify limits on particular disclosures. But confidentiality should not become a blanket reason for withholding any meaningful account. Regulators and reviewing bodies may need access to more technical information than the affected individual, while the individual should still receive an explanation sufficient to understand and contest the outcome.

The central test is therefore not whether an organisation has made its algorithm publicly transparent. It is whether a person affected by an AI-assisted decision can understand the decision’s relevant grounds, identify a possible error and access an effective review process.

Digital governance should not turn consequential decisions into unanswerable outcomes. When public authorities or private organisations rely on AI in ways that significantly affect individuals, explanation, human oversight and effective avenues of challenge should operate as connected safeguards. The right to explanation is not a demand to reveal every technical secret; it is a demand that decision-making power remain intelligible and accountable to the people subject to it.

The views expressed are the author’s own.

AI Assistance Disclosure:

Generative AI tools were used to assist with language refinement and editorial revisions. The author reviewed the revised text and remains responsible for its content, arguments, and conclusions.