The Right to Algorithmic Explanation: Legal Foundations

This page is a translation of the original Portuguese text, provided for convenience. It deals with Brazilian law. If there is any discrepancy, the Portuguese version prevails. Read the original in Portuguese

The right to algorithmic explanation is emerging as a fundamental safeguard for individuals facing automated decisions that affect their rights, freedoms, and legitimate interests.

What the Right to Algorithmic Explanation Is

We live in an era in which algorithms make decisions that directly affect people’s lives. Automated systems operate at scale, from approving or denying credit to screening résumés in hiring processes, from reviewing social security benefits to building risk profiles for insurance. Often, the people affected have no idea what criteria produced the result. In this setting, the right to algorithmic explanation is an essential legal tool for ensuring transparency, oversight, and accountability in the relationship between individuals and artificial intelligence systems.

When we talk about the right to an explanation for algorithmic decisions, we mean every person’s entitlement to clear, accessible, and sufficient information about how an automated decision was made: which data were considered, which logic was applied, and which criteria determined the final outcome. It is not enough to know that an algorithm made the decision. You are entitled to understand why it decided the way it did. That understanding is a prerequisite for exercising other rights, such as the right to be heard, the right to a full defense, and the right to have potentially discriminatory or mistaken decisions reviewed.

This right is part of a global movement to regulate artificial intelligence, one that recognizes the information gap between the developers of automated systems and the individuals affected by their decisions. Without mechanisms for explainability, people are held hostage by a technological “black box.” They cannot challenge unfair results or exercise any control over processes that affect their legal rights.

Constitutional and Statutory Foundations in Brazilian Law

Brazil’s 1988 Federal Constitution provides a solid basis for the right to algorithmic explanation, even though it does not mention the right expressly. The principle of human dignity, one of the foundations of the Republic, requires that individuals not be treated as mere objects of data processing. They must be treated as subjects with autonomy and the capacity to understand the decisions that affect them. Likewise, the constitutional right to information supports the demand for transparency in automated decision-making, especially when it is carried out by public agencies or by entities that perform functions of collective interest.

Due process of law, in both its procedural and substantive dimensions, also supports the demand for explainability. If an administrative or judicial decision is made with the help of algorithmic systems, the rights to be heard and to a full defense presuppose that the person affected understands the grounds for the decision. An opaque decision, whose criteria remain out of reach, directly violates those procedural guarantees.

At the statutory level, Brazil’s General Data Protection Law (LGPD), Law No. 13.709/2018, is the most important legal framework on the subject. Article 20 of the LGPD gives data subjects the right to request review of decisions made solely through automated processing of personal data that affect their interests, including decisions meant to define their personal, professional, consumer, or credit profile, or aspects of their personality. The original text expressly provided that this review had to be carried out by a natural person, but a later amendment relaxed that requirement. That change has prompted significant debate among legal scholars about how effective the provision really is.

Beyond the right to review, the LGPD establishes, in Article 6, principles that reinforce the demand for explainability. The transparency principle guarantees data subjects clear, precise, and easily accessible information about how their data are processed. The purpose principle limits processing to legitimate, specific purposes that have been disclosed to the data subject. The non-discrimination principle prohibits processing for unlawful or abusive discriminatory purposes. All of these principles point to the same conclusion: algorithmic decisions must be explainable and open to challenge.

Algorithmic explainability is not a technological luxury. It is a democratic requirement, so that individuals can exercise control over automated decisions that affect their fundamental rights.

We should also note that the Marco Civil da Internet (Brazil’s Internet Civil Rights Framework), Law No. 12.965/2014, and its implementing decree add to this body of law. They establish principles of transparency and user protection in digital environments. Taken together, these rules create a system of protection that, although still taking shape, offers concrete tools for enforcing the right to algorithmic explanation.

AI Regulation and the Brazilian Bill

The debate over the right to algorithmic explanation has taken on a new dimension as Brazil’s legislative process to regulate artificial intelligence moves forward. Bill No. 2338/2023, which moved through the Federal Senate and proposes a Legal Framework for Artificial Intelligence, pays specific attention to the transparency and explainability of AI systems. It classifies those systems according to the degree of risk they pose to fundamental rights.

The Brazilian proposal draws on international experience, most notably the European Union’s AI Act (Regulation 2024/1689), which sets graduated transparency obligations based on an AI system’s risk classification. High-risk systems, such as those used for credit decisions, hiring, educational assessment, and law enforcement, face stricter requirements for documentation, explainability, and human oversight.

In the Brazilian context, the bill provides that high-risk AI systems must give affected individuals information about the logic and criteria behind an automated decision, in accessible and understandable language. This requirement goes beyond simply disclosing that an algorithm was used. The explanation must be substantive. It must let the person affected truly understand the factors that determined the outcome and, where appropriate, contest it on an informed basis.

Sector-specific regulation is also advancing. In the financial sector, the Central Bank of Brazil and the Brazilian Securities and Exchange Commission have issued rules requiring transparency in algorithmic models for credit analysis and risk management. In health care, the National Health Surveillance Agency is discussing regulatory frameworks for software as a medical device, including explainability requirements. In labor law, there is growing discussion of the need for transparency in the algorithms that manage digital platforms and shape the working conditions of millions of people.

Technical and Legal Challenges of Explainability

Putting the right to algorithmic explanation into practice faces significant challenges at the intersection of law and technology. The first is the tension between explainability and computational complexity. Many of the most advanced artificial intelligence systems, such as deep neural networks (deep learning), are inherently opaque. It is difficult, and sometimes technically infeasible, to give complete and accurate explanations of how they reached a particular result. This feature, known as the “black box” problem, is a real obstacle to enforcing the right.

Legal scholars and the technical community have developed approaches to ease this problem. Explainable AI (XAI) techniques, such as LIME (Local Interpretable Model-agnostic Explanations) and SHAP (SHapley Additive exPlanations), generate interpretable approximations of complex models’ decisions by identifying which variables carried the most weight in a given result. These explanations are simplifications, but they can still meet the legal standard of explainability. That standard does not require a full reproduction of the computational process. It requires an understanding sufficient to exercise one’s rights.

Another important challenge is defining the right standard of explanation. Different contexts call for different levels of detail. An explanation addressed to a data subject who is a layperson must be accessible and easy to understand. One addressed to a regulator or to the judiciary can and should be more technical and in-depth. The law needs to account for these gradations. It should avoid both explanations so simplified that they become useless and demands for technical detail that ordinary people cannot make sense of.

Protection of trade secrets and the intellectual property in algorithms is another point of tension. AI developers often argue that disclosing the criteria and logic of their algorithms would compromise their trade secrets. That argument has some legitimacy, but it cannot prevail over individuals’ fundamental rights in an absolute way. The solution lies in building mechanisms that allow a functional explanation (what the algorithm does and why it reached a given result) without necessarily revealing the source code or proprietary implementation details.

The problem of algorithmic discrimination also deserves attention. AI systems can reproduce and amplify biases present in their training data, producing discriminatory results based on race, gender, age, origin, or socioeconomic status. In this context, the right to algorithmic explanation works as a tool for detecting and correcting those distortions. Without explainability, it is practically impossible to identify and prove that an automated decision was discriminatory, which makes anti-discrimination protection unworkable.

Practical Applications and Outlook in Brazilian Law

The practical importance of the right to algorithmic explanation shows up in situations that are increasingly common in Brazilian legal practice. In social security, for example, the INSS (Brazil’s National Social Security Institute) has invested in automated systems to review benefit applications. When a retirement or benefit claim is denied on the basis of algorithmic criteria, the insured person has the right to understand which data were considered, which rules were applied, and why the outcome was unfavorable. Without that understanding, an administrative or judicial appeal becomes an exercise in the dark.

In consumer law, the use of algorithms for pricing (dynamic pricing), credit analysis, and consumer profiling raises direct questions of explainability. The Consumer Protection Code guarantees the right to clear and adequate information about products and services. That guarantee provides additional support for requiring that algorithmic decisions affecting consumer relationships be explainable.

In labor law, the algorithmic management of workers on digital platforms (such as ride-hailing and delivery apps) has produced litigation in which explainability is a central issue. Workers who are deactivated or penalized by automated systems often do not know the criteria behind the decision, which leaves them unable to mount any effective defense. Case law in this area is likely to establish algorithmic transparency as part of the duty of good faith in contracts.

The Brazilian judiciary also faces the challenge of explainability in its own systems. Artificial intelligence tools used to screen cases, suggest decisions, and calculate sentences must be transparent and open to scrutiny. Otherwise, due process guarantees are at risk. Resolution No. 332/2020 of the National Council of Justice set guidelines for the ethical use of AI in the judiciary, including requirements for transparency and explainability.

In conclusion, the right to algorithmic explanation is an essential frontier of legal protection for individuals in the digital society. Making it effective will take a joint effort by legislators, regulators, legal professionals, and technology developers to build standards of explainability that are technically feasible, legally sound, and accessible to the public. The legal foundation already exists in Brazilian law. The challenge now is to turn it into concrete, effective practice.

As informações deste artigo são de caráter informativo e não substituem consulta jurídica individualizada. Cada caso possui particularidades que exigem análise específica.