Design Principles and Ethics

The TANGO project embeds ethical considerations and robust design principles directly into its methodology, ensuring its systems are trustworthy and human-centric. The framework is explicitly assessed against the four core principles for ethical AI: respect for human autonomy, prevention of harm, fairness and explicability. A central design goal is to enhance, not undermine, human decision-making capabilities, allowing users to make informed decisions without being subjected to undue nudging.

Compliance with legal and regulatory standards is a cornerstone of the project. The methodology ensures adherence to existing EU regulations such as the General Data Protection Regulation (GDPR) and the e-Privacy Directive, and it is designed to be future-proof by considering forthcoming legislation like the AI Act, the Data Act and the Data Governance Act. The "Privacy by Design" principle is applied throughout, with data protection impact assessments performed where necessary to safeguard personal and sensitive data.

Beyond legal compliance, TANGO adopts a "data justice" approach, which prioritizes equity, recognition of plural interests, and the creation of public goods. This is coupled with the "do no significant harm" principle, ensuring the project's activities do not negatively impact environmental objectives. A major focus is the mitigation of algorithmic bias. The project actively develops methods to audit models for unfair treatment of subgroups based on attributes like gender or race and to counteract historical injustices that may be present in training data.

The design emphasizes rich, human-computer interaction (HCI) strategies. This includes mixed-initiative interaction, a flexible dialogue where human and AI agents can each contribute what they do best, and argumentative decision-making, which encourages partners to formalize their reasoning to reach a shared, well-supported conclusion. Furthermore, the system supports anticipatory decision-making, particularly for algorithmic recourse, where it interactively provides users with interpretable and actionable steps to overturn an unfavorable automated decision. This interdisciplinary fusion of cognitive science, law and computer science aims to establish a new governance paradigm for AI-assisted decision-making.