Stakeholders map
An overview of key actors, their roles, incentives, decision authority, and interactions across the project lifecycle.
Human-Oriented AI: A Design Framework
Socio-technical framework for the responsible design and deployment of AI-based systems in the public sector.
Competency Definition
Context diagram
A high-level map of the institutional, service, data, and technology environment in which the AI initiative operates.
Public value canvas
A structured canvas that links societal goals, service outcomes, beneficiaries, implementation choices, and evaluation criteria.
Principles catalog
A curated set of design and governance principles that guide responsible choices across strategy, development, deployment, and oversight.
Requirements catalog
A structured catalog of functional, organizational, legal, and operational requirements to guide responsible AI design and implementation.
Risk catalog
A consolidated view of technical, ethical, legal, and organizational risks, including mitigation strategies, ownership, and monitoring actions.
Policy and regulatory map
A synthesis of relevant legal, policy, and governance requirements that shape design and deployment choices.
Public value goals and outcomes
Defining clear public value objectives for AI initiatives and aligning project design choices with societal, institutional, and service goals.
Data requirements
Defining the data specifications needed for model performance, governance compliance, and meaningful use in context.
Data availability
Identifying and documenting the data sources currently available for model development, validation, and operational use.
Data accessibility
Ensuring authorized stakeholders can discover, access, and use data through clear governance rules, permissions, and usable access mechanisms.
Data quality
Assessing completeness, consistency, representativeness, and reliability of datasets to support trustworthy AI outcomes.
Model selection
Choosing model approaches that fit the problem, available data, explainability needs, and operational constraints.
Performance metrics
Selecting technical and service-level metrics to evaluate model quality, reliability, and expected public value outcomes.
Model constraints
Specifying limits related to fairness, interpretability, resources, and policy requirements that shape acceptable model behavior.
Computing resources
Planning processing capacity, storage, and supporting infrastructure required across development and production stages.
Deployment environment
Defining target environments, integration dependencies, and operational interfaces for reliable deployment.
Scalability needs
Estimating growth in users, data volume, and service demand to design systems that scale sustainably.
Performance monitoring
Establishing continuous monitoring for model quality, service performance, and drift in operational conditions.
Model updates
Designing controlled update processes for retraining, validation, approval, and safe release of model changes.
Maintenance plan
Defining long-term ownership, support routines, incident handling, and lifecycle maintenance responsibilities.
Anticipating impact
Anticipating potential consequences of AI use in public services before deployment, including social, institutional, and rights-related effects.
Landscape of data ethics frameworks
Understanding major ethics frameworks and applying them pragmatically to support responsible decisions in specific policy contexts.
Reflecting on causes and mechanisms
Examining assumptions, institutional drivers, and socio-technical dynamics that shape AI outcomes in public-sector settings.
Deliberating options
Comparing alternative design and governance options through inclusive discussion, balancing trade-offs among values, risks, and feasibility.
Embedding responsiveness
Implementing feedback, monitoring, and adaptation mechanisms so governance can respond to evolving impacts after deployment.