Results
The AI4Gov project has delivered a comprehensive set of results designed to promote ethical, transparent, and citizen-centric use of AI in public governance.
Frameworks
Holistic Regulatory Framework
The Holistic Regulatory Framework (HRF) developed is a comprehensive framework designed to regulate the use of AI technologies in governance. Its primary purpose is to ensure that AI systems used in public services are developed, deployed, and operated in a manner that is ethical, transparent, and compliant with applicable laws and regulations such as the General Data Protection Regulation (GDPR) addressing in parallel the issues of bias and discrimination. The HRF also incorporates ethical recommendations from various AI bodies, including the High-Level Expert Group on Artificial Intelligence (HLEG) and aligns with regulations such as the General Data Protection Regulation (GDPR) and the AI Act.
Data Governance Framework
The Data Governance Framework (DGF) is a structured and comprehensive set of guidelines, policies, and procedures that govern how data is managed, shared, and protected within the AI4Gov Project. This framework serves as a crucial instrument for ensuring that data-related activities align with the EU's legal and regulatory landscape, particularly with regard to data protection and privacy. Within this context, the Data Governance Framework project plays a pivotal role in navigating the complexities of data management while complying with EU data protection laws. This framework acts as a structured roadmap that not only empowers project partners to harness the potential of data but also safeguards the rights and interests of individuals whose data is processed.
Technical tools
Virtualized Unbiasing Framework for AI & Big Data
The VUF is a suite of AI-driven analytics techniques designed to create and/or enhance policies based on predictive analytics results and provide a set of citizen-centric functionalities. All functionalities are accessible via an intuitive visualization environment and can be dynamically invoked, reused, and adapted for diverse datasets and heterogeneous data sources. It consists of the following components:
Bias Detector Toolkit
The Bias Detector Toolkit is a holistic application focused on explaining AI bias and equipping developers with an easy-to-navigate and visually organized catalogue. It consists of the scrollytelling application to explain bias and AI, real life examples and the catalogue of methods and tools for bias mitigation.
Adaptive Analytics Framework
The Adaptive Analytics Framework is the component that develops the needed ML models for performing predictive analytics and optimised resource allocation to satisfy the needs of the pilots and assist policy makers.
Policy-Oriented Analytics and AI Algorithms
The Policy-Oriented Analytics and AI Algorithms is the component that develops several NLP algorithms in order to analyse large volumes of text data and also assist the respective AI experts.
Visualization Workbench
The Visualization Workbench is an interactive platform that enables users to explore, analyze, and interpret complex datasets through intuitive visual representations. It supports dynamic data integration, customizable dashboards, and advanced analytics, making it easier for policymakers and stakeholders to gain actionable insights and foster evidence-based decision-making.
Policy Recommendation Toolkit
The Policy Recommendation Toolkit aims to facilitate organizations in policy-making and provide a transparent governance model that gives the opportunity to citizens to audit these processes of policy-making and to actively participate in the formation of them via blockchain-enabled co-creation. Building on this foundation, the Toolkit incorporates data-driven analysis and structured decision-support tools that help policymakers evaluate information and compare alternative policy scenarios. It enables experts, stakeholders, and citizens to contribute insights through an accessible participatory environment while maintaining clear accountability. By leveraging blockchain technology, all interactions - such as policy creation, vote by citizens, and endorsements -are securely recorded and verifiable, ensuring transparency, immutability and integrity throughout the entire process. This creates a trusted ecosystem where policies are developed collaboratively, monitored effectively, and continuously refined to better reflect societal needs.
Citizen’s Wallet
The Citizen’s Wallet is a secure, wallet-based toolkit that enables encrypted voting, identity management, and verifiable credentials. It leverages cryptographic signing and decentralized identifiers to ensure privacy, trust, and interoperability for citizens and policymakers.
Blockchain-based tools / Blockchain-based Information Exchange (BIE)
The BIE is a holistic solution for regulating access to the data by the various participants and facilitating the secure & trustful exchange of data across all stakeholders.
Virtualized Unbiasing Framework for AI & Big Data
The VUF is a suite of AI-driven analytics techniques designed to create and/or enhance policies based on predictive analytics results and provide a set of citizen-centric functionalities. All functionalities are accessible via an intuitive visualization environment and can be dynamically invoked, reused, and adapted for diverse datasets and heterogeneous data sources. It consists of the following components:
Bias Detector Toolkit
The Bias Detector Toolkit is a holistic application focused on explaining AI bias and equipping developers with an easy-to-navigate and visually organized catalogue. It consists of the scrollytelling application to explain bias and AI, real life examples and the catalogue of methods and tools for bias mitigation.
Adaptive Analytics Framework
The Adaptive Analytics Framework is the component that develops the needed ML models for performing predictive analytics and optimised resource allocation to satisfy the needs of the pilots and assist policy makers.
Policy-Oriented Analytics and AI Algorithms
The Policy-Oriented Analytics and AI Algorithms is the component that develops several NLP algorithms in order to analyse large volumes of text data and also assist the respective AI experts.
Visualization Workbench
The Visualization Workbench is an interactive platform that enables users to explore, analyze, and interpret complex datasets through intuitive visual representations. It supports dynamic data integration, customizable dashboards, and advanced analytics, making it easier for policymakers and stakeholders to gain actionable insights and foster evidence-based decision-making.
Policy Recommendation Toolkit
The Policy Recommendation Toolkit aims to facilitate organizations in policy-making and provide a transparent governance model that gives the opportunity to citizens to audit these processes of policy-making and to actively participate in the formation of them via blockchain-enabled co-creation. Building on this foundation, the Toolkit incorporates data-driven analysis and structured decision-support tools that help policymakers evaluate information and compare alternative policy scenarios. It enables experts, stakeholders, and citizens to contribute insights through an accessible participatory environment while maintaining clear accountability. By leveraging blockchain technology, all interactions - such as policy creation, vote by citizens, and endorsements -are securely recorded and verifiable, ensuring transparency, immutability and integrity throughout the entire process. This creates a trusted ecosystem where policies are developed collaboratively, monitored effectively, and continuously refined to better reflect societal needs.
Blockchain-based tools / Blockchain-based Information Exchange (BIE)
The BIE is a holistic solution for regulating access to the data by the various participants and facilitating the secure & trustful exchange of data across all stakeholders.
Citizen’s Wallet
The Citizen’s Wallet is a secure, wallet-based toolkit that enables encrypted voting, identity management, and verifiable credentials. It leverages cryptographic signing and decentralized identifiers to ensure privacy, trust, and interoperability for citizens and policymakers.
Situation-aware Explainability (SAX4BPM)
The SAX4BPM library is an open-source Python library developed by IBM Research that provides tools for deriving explanations about business processes (BPM) by integrating process mining, causal discovery, and Explainable AI (XAI) with Large Language Models (LLMs). The library's primary goal is to generate "Situation-aware Explanations" (SAX explanations) to help users understand why certain conditions or outcomes occur within a business process, thereby promoting trust in AI-augmented BPM systems.
- Deliverable 4.1 – Trustworthy, Explainable, and unbiased AI V1 (SAX4BPM architecture and services)
- (constructs. results and implications of SAX4BPM user study)user study)
- SAX4BPM is available here
- Documentation: SAX4BPM
Other Tools
Training & Learning materials
AI4Gov has created a set training and learning resources, including MOOCs, aimed at enhancing awareness and citizen engagement, participation and inclusiveness in democratic procedures, as well as raising awareness and knowledge about bias in AI. A Guideline for National Governments/Ministries has also been developed, that outlines a strategy for fostering AI Literacy through Informal Learning & Open Education.
Self-assessment tools
AI4Gov has developed a set of practical self-assessment tools, aimed in ensuring compliance of AI systems with ethical principles and legal regulations.
