Project Information

Supported by European Union’s Research and Innovation framework programme, Horizon 2020, AI4PublicPolicy is a joint effort of policy makers and Cloud/AI experts to unveil AI’s potential for automated, transparent and citizen-centric development of public policies. To this end, the project will deliver, validate, demonstrate and promote a novel open cloud platform, the AI4PublicPolicy Platform, for automated, scalable, transparent and citizen-centric policy management based on unique AI technologies.

The AI4PublicPolicy platform will be an open Virtualized Policy Management Environment (VPME) that will provide fully-fledged policy development/management functionalities based on AI technologies such as Machine Learning (ML), Deep Learning (DL), NLP and chatbots, while leveraging citizens’ participation and feedback. It will support the entire policy development lifecycle, based on technologies for the extraction, simulation, evaluation and optimisation of interoperable and reusable public policies, with emphasis on citizen-oriented feedback loops.

AI4PublicPolicy will complement public policy development functionalities ensuring the effective transition from legacy policy development models to emerging AI-based policymaking.

The AI4PublicPolicy VPME will be integrated with the European Open Science Cloud (EOSC) platform with a dual objective. First to facilitate access to the Cloud and high-performance computing (HPC) resources of EOSC/EGI that are required to enable the project’s AI tools; and second to boost the sustainability and wider use of the project’s developments.

AI4PublicPolicy’s business plan for sustaining, expanding and commercialising the AI tools and the VPME is based on the development of a community of interested and engaged stakeholders (i.e. public authorities and other policymakers) around the project’s platform.

The Challenge

Policy development represents one of the most prominent applications of Cloud-computing and HPC for public administrations at local, regional and national levels, for harvesting vast amounts of data, including governmental databases and interactions with the citizens, data from public infrastructures, as well as from alternative sources such as social networks and the public internet, while improving the scalability, cost-efficiency, flexibility and quality of public services.

The use of AI-based solutions for data-driven policymaking can address technical and organisational challenges such as:

• Access to Cloud and HPC resources: Scalable infrastructures for storing and analysing datasets are not readily available within public authorities.
• Transparency: Several AI systems operate as “black-boxes”, making their use challenging, since policymaking must be transparent in terms of the rationale behind the proposed policies and the factual data that drive policy developments.
• Trustworthiness: AI systems for policymaking must be secure, safe and resilient against malicious attempts, as well as address the ethical AI challenges, as analysed by EU’s High-Level Expert Group (HLEG) on Artificial Intelligence.
• End-to-End Integration: AI’s potential in policymaking is significantly under-exploited, given it provides opportunities for automating and supporting policymaking activities, including front office, middle office, and back-office activities.
• Organisational Transformation: The introduction of AI systems in policymaking leads to automation and efficiency, yet it challenges peoples’ roles and responsibilities, requiring significant upskilling and reskilling for policy making experts and other employees of the public administration.

Objectives

Methodology

The AI4PublicPolicy workplan is deployed in three progressive phases

Phase 1 | Specification & Fine Tuning of the AI4PublicPolicy Concept (M1-M9)

The project starts off with the collection of insights based on user studies involving public authorities’ personnel and policymaking stakeholders, and the review of reference scenarios for evidence-based policy development.

To ensure up to date research outcomes, the project performs an analysis on the state-of-the-art regarding main technical areas and complements it with relevant standards, applicable regulations, and project-specific requirements.

The derived knowledge is used to create a set of requirements and “innovation forms”, each split into prioritised functionalities to articulate, fine-tune and specify the project in detail.

Phase 1 concludes to the specification of the reference architecture and organisational transformation blueprints for AI-based policy development to steer the technological WPs’ processes and specify the AI4PublicPolicy Requirements and Innovation tracker.

Phase 2 | Initial Integration & Technical Validation (M10-M24) 

The initial Proof-of-Concept prototypes are developed (i.e., Most Viable Poroducts – MVP), along with pilot systems and use cases, which are tested based on the initial release of AI4PublicPolicy VPME and AI Tools.

Project’s prototypes are tested in partners’ pilot sites based on real-life datasets and policymaking use cases.

Driven by the delivered specifications, Phase 2 produces the initial versions of project’s (i) Semantic interoperability for policies and datasets, (ii) Transparent and Trustworthy AI solutions, and (iii) AI Tools for policymaking and citizen-centric optimization.

During this phase, the first iteration of VPME is integrated to drive the implementation of pilots and use cases; and is validated through appropriate user studies and evaluation workshops.

Phase 3 | Technical & Business Validation (M24-M36)

The project’s prototypes are revised through (a) functionality enhancements with advanced versions of the algorithms and components that support their operation and (b) fine-tuning of the prototypes based on stakeholders’ feedback.

The second iteration of the prototypes of project’s solutions is delivered to enable the production of the second integrated version of the project’s VPME platform, followed by the second cycle of pilot deployments and operations.

Phase 3 entails the initial business and socio-economic validation based on additional user studies and in conjunction with the project’s business modelling activities.