Ethics and Artificial Intelligence: Transparency (4/7)
Principles for trustworthy artificial intelligence
In 2019, the European Commission's High-Level Expert Group on Artificial Intelligence (AI HLEG) published its "Ethics Guidelines for Trustworthy AI". To this day, it remains one of the foundational documents on the ethical requirements for artificial intelligence.
The Guidelines set out seven key requirements for trustworthy AI that must be met at every stage of a system's lifecycle - from design and development to deployment and real-world use.
These requirements are:
-
Human agency and oversight - read our article here
-
Technical robustness and safety - read our article here
-
Privacy and data governance - read our article here
-
Transparency
-
Diversity, non-discrimination and fairness - read our article here
-
Societal and environmental well-being - read our article here
-
Accountability - read our article here
In this post we look at the fourth principle: transparency.
What does the principle of transparency mean?
Transparency in artificial intelligence (AI) means clarity and openness about how algorithms work, what data they use, and how they make decisions. It is not merely the explainability of individual outputs, but comprehensive information about the design, data, processes, and purposes of the system.
Transparency in the development and use of AI systems is critically important for the following reasons:
-
Opening the "black box" Many AI systems operate as untraceable "black boxes" whose internal logic is hard to understand. Transparency lets us open that box and grasp the mechanisms behind the system's decisions.
-
Trust, control, and compliance Amid growing regulation, including within the European Union, transparency is a key element in building trust and ensuring compliance with the law.
-
Adaptability and accountability Transparency is not just a technical feature. It involves data governance, clearly defined objectives, ongoing dialogue with stakeholders, and the provision of information at different levels - from users to oversight bodies.
Three dimensions of transparency
According to the Ethics Guidelines for Trustworthy AI, the principle of transparency in the context of trustworthy artificial intelligence has three main dimensions.
Traceability
Traceability means that the data and processes that led to a given decision by an AI system must be documented. This covers both the data sources and how they were labelled, as well as the algorithms that were used. Specifically, it requires:
-
knowing what data was used to train the model;
-
keeping a record of which algorithms were applied;
-
storing the intermediate stages of processing.
Goal: to identify the causes of errors and prevent future inaccuracies.
Explainability
Explainability refers to a system's ability to provide understandable reasons for its decisions. This does not necessarily mean disclosing the entire source code, but rather enabling humans to understand the underlying logic.
Goal: for AI decisions to be understood and verifiable.
Communication
Communication concerns clearly indicating that a person is interacting with an AI. This ensures:
-
that users are not misled into believing they are talking to a human;
-
that they are informed about the system's role in the decision-making process;
-
that they know where the limits of responsibility lie.
Goal: for people to be informed that the results they receive come from an AI system, and to understand the logic behind them.
Transparency in practice
Transparency delivers real value when it is built into the everyday work with artificial intelligence systems. This requires a deliberate choice of technologies that clearly show how they function, along with concrete practices on the part of organizations. In practice, transparency can be achieved in many ways, including:
-
Asking questions Users should be able to ask: "What is this answer based on?", "Which sources were used?". AI systems should provide understandable explanations and references to the data.
-
Documenting processes Organizations should maintain clear internal policies for deploying AI, as well as a register of the data and algorithms used. This makes oversight easier and prevents the risk of "black boxes".
-
Informing users Everyone should know when they are interacting with an AI, what the system's capabilities and limitations are, and how far its responsibility extends.
Transparency is not just good practice - it is a selection criterion. Systems like Praven Intelekt are built to show what data and logic their answers are based on. This guarantees trust, security, and control, which is especially important amid rapidly evolving technologies and strict regulatory requirements.
An example of how Praven Intelekt lets you easily see all the documents an answer is based on, including references to the specific article and paragraph in the sources.
