MARTA

Machine Learning Auditing for Robust and Transparent Administration

 

Description:

As Artificial Intelligence becomes integral to modern public services, ensuring its ethical and trustworthy application is more urgent than ever. The MARTA project addresses this need by developing robust auditing processes and governance best practices tailored for AI systems in public administration.

The project focuses on the “e-Balcão Inteligente” service of the Portuguese Tax Authority (Autoridade Tributária - AT), where AI models assist in classifying and responding to taxpayer inquiries. While these models enhance efficiency, they also introduce risks related to bias, privacy, and transparency. MARTA aims to audit and improve these systems by creating tools that assess fairness, robustness, and data quality, ensuring that AI decisions are accurate, explainable, and compliant with regulatory standards.

Beyond this use case, MARTA’s findings will feed into the AICeBlock platform, a blockchain-based framework for AI certification, supporting a standardised and scalable approach to responsible AI development across the public sector.

By aligning with European regulations and engaging stakeholders from government to academia, MARTA is establishing a new standard for AI governance, transparency, and accountability in public services.

 

Consult the project's Spec Sheet here:

English

 

Funded by:

For any additional information, please contact us using the inquiries form.