Hierarchical HCAI: A Multilevel Framework

• Hierarchical HCAI is a multi-level framework that embeds human authority, values, and accountability into every layer of sociotechnical systems.
• It replaces siloed approaches with joint optimization methods that align technical performance with ethical, cultural, and organizational imperatives.
• The framework employs nested design hierarchies and feedback mechanisms to guarantee transparent inter-level governance and continuous co-evolution between AI and human roles.

Hierarchical Human-Centered AI (hHCAI) is a paradigmatic extension of human-centered artificial intelligence that situates human needs, values, and ultimate authority at the apex of AI system design across multiple, nested levels of sociotechnical context. Explicitly grounded in both traditional sociotechnical systems (STS) theory and the methodology of human-centered AI (HCAI), hHCAI replaces siloed, application-specific approaches with a structured, multilevel framework that operationalizes human-centered principles at the scale of individuals, organizations, ecosystems, and macro-social systems. The approach formalizes joint optimization between technical and social subsystems, emphasizing co-evolution, accountability, situational awareness, and governance along every layer, thereby ensuring AI technologies remain operationally aligned with ethical, cultural, policy, and organizational imperatives (Xu et al., 2024, Gao et al., 16 Jan 2026, Xu et al., 2023).

1. Foundational Framework and Definitions

hHCAI is defined as a multi-level, interdisciplinary methodology for the design, development, and deployment of AI systems such that human authority, skill, and values are not only preserved but amplified. Within the iSTS (intelligent sociotechnical systems) framework, hHCAI functions as the central design engine driving joint optimization at every hierarchical layer, extending earlier STS work by:

• Elevating AI systems from passive tools to collaborative teammates.
• Embedding learning, autonomy, and dynamic co-evolution with humans and social structures.
• Expanding design scope from individuals and organizations to intelligent ecosystems and entire societies (Xu et al., 2024, Gao et al., 16 Jan 2026).

The core construct is a layered sociotechnical system, where at each layer $\ell \in \{1, 2, 3, 4\}$, the composite performance $P_\ell$ is a function of its technical subsystem $T_\ell$, social subsystem $S_\ell$, and coupling constraints $C_\ell$ inherited from its parent layer:

$P_\ell = F_\ell(T_\ell, S_\ell, C_\ell)$

These coupling constraints capture how broader organizational, ecosystemic, or societal values delimit and shape the optimization of any given layer (Xu et al., 2024).

2. Hierarchical Structure and Inter-Level Dynamics

Levels of hHCAI

hHCAI is characterized by explicit, nested layers—sometimes metaphorized as rings or the "point–plane–body" progression (Gao et al., 16 Jan 2026, Xu et al., 2023):

Layer	Tenor	Representative Contexts	Design Focus
Individual	"Human-in-the-loop"	Human–AI joint cognitive systems (HAII, e.g., driver & vehicle)	Explainability, final human control, shared SA, dynamic teaming
Organization	"Organization-in-the-loop"	Work systems, operating units, flight departments	Role/process redesign, governance, standardization
Ecosystem	"Ecosystem-in-the-loop"	Multi-organization, V2V/V2I networks, smart cities	Coordination, ethical alignment, resource-sharing
Society	"Society-in-the-loop"	Regulation, policy, broad cultural values	Law, ethics, sustainable governance, value alignment

At each layer, the organizational logic is both serving upwards (realizing goals from the level above) and constraining downwards (translating context-appropriate requirements for layers below).

Interactions and Feedback

• Nested constraint flows: Societal policies constrain ecosystem design; emergent behaviors at the ecosystem level inform organizational protocols, which refine individual H–AI teaming (Gao et al., 16 Jan 2026).
• Feedback mechanisms: User experience and trust metrics at the individual level feed back into organizational governance and societal policy evolution.
• Progressive scale: Implementation typically proceeds from micro (pilot deployments, joint cognitive teaming) through meso (organizational/ecosystem integration) to macro (embedding in regulation and public deliberation).

3. Updated Sociotechnical Design Principles

hHCAI articulates a revised set of ten sociotechnical design dimensions for the AI era, adapted from STS theory to integrate autonomy, learning, and ecosystem-level interdependence (Xu et al., 2024). These include:

• Transitioning machines from passive tools to full collaborative teammates.
• Redefining the human–machine relationship as partnership and co-evolution rather than operator–artifact.
• Expanding design scope from user interfaces to include organizational reporting chains, ecosystem interoperability, and external governance.
• Elevating learning ability as continuous, networked, and distributed across human and AI agents.
• Making organizational and societal goals explicit design constraints, not ex post factors.
• Integrating monitoring, diagnosis, coordination, and compensation cycles for dynamic sociotechnical adaptation.

Across individual, organizational, ecosystem, and societal levels, these principles direct how roles, authority, workflow, governance, and situational awareness are rearchitected to support genuine human-AI joint optimization (Xu et al., 2024).

4. Methodological Taxonomy and Implementation

hHCAI prescribes both a requirements hierarchy and practical method taxonomy for end-to-end human-centered AI development (Xu et al., 2023, Gao et al., 16 Jan 2026).

Requirements Hierarchy

A six-level decomposition governs translation from principle to execution:

Level	Artifact Type	Example(s)
ℓ=0	Design Philosophy	"AI must serve humans, maximizing benefit and minimizing harm."
ℓ=1	Design Goals	Trustworthy, scalable, empowering, usable, responsible, etc.
ℓ=2	Design Principles	"Ensure human ultimate authority," "Enable explainable AI," etc.
ℓ=3	Implementation Approaches	Explainable AI, hybrid human-in-the-loop intelligence
ℓ=4	Methods	HCAI-centric ML, participatory UX, governance mechanisms
ℓ=5	Processes	Double Diamond design, AI lifecycle integration

Mappings are explicitly modeled via binary matrices (e.g., Mapping$_{GP}$, Mapping$_{PA}$, etc.), ensuring traceability and methodical coverage throughout process execution (Xu et al., 2023).

Taxonomy of Methods

hHCAI groups implementation techniques into five broad categories (Gao et al., 16 Jan 2026):

• Human-centered strategies: Value alignment, hybrid intelligence, data/knowledge fusion.
• Computation and modeling: Human-in-the-loop ML, explainable AI, joint cognitive architectures.
• Human controllability: Final human authority, override/controllability, safe fail-over.
• Interaction design: Shared situation awareness interfaces, adaptive modalities.
• Standards and governance: Ethics-by-design, audit frameworks, accountability protocols.

These categories span strategic, technical, experiential, and regulatory dimensions, each closely mapped to corresponding hierarchical layers.

5. Concrete Domains and Case Studies

Empirical studies establish the practical salience of hHCAI across transportation, aviation, and allied sectors (Gao et al., 16 Jan 2026, Xu et al., 2023):

• Autonomous Driving: At the individual level, the vehicle functions as a cognitive collaborator employing dual-layer situation awareness. At the ecosystem level, fleets interoperate via V2V/V2I protocols. Societal layer governance imposes standards for transparency and responsibility. Empirically, exposing AI’s situational predictions improves human driver trust calibration and system reliability.
• Single-Pilot Cockpit Operations: Cockpit AI partners with the lone pilot via joint decision-making and situation awareness modules; override authority and predictability are paramount. Ecosystem and societal levels introduce coordination with air traffic control and regulatory compliance, with public safety and ethical mandates reflected in cockpit system design.

These cases evidence how requirements and principles percolate from abstract philosophy to concrete interface features, role allocations, and governance structures.

6. Addressing Limitations and Systemic Challenges

hHCAI directly addresses recognized limitations of classical HCAI, which often restrict focus to individual explainability or ethical alignment while neglecting systemic organizational, ecosystemic, or societal impacts (Xu et al., 2024, Xu et al., 2023):

• Overcoming narrow scope: By explicitly extending the "loop" to organizations, ecosystems, and society, hHCAI attends to the broader consequences and cross-actor dependencies of AI deployment.
• Dynamic co-evolution: Positioning AI agents as collaborative teammates (rather than tools) enables mature teamwork theories, continuous co-learning, and dynamic adaptation.
• Integrated governance: Rather than retrofitting policy post-deployment, hHCAI prescribes governance and design processes that cascade from society down to individual workflows, ensuring systemic, not local, alignment.
• Actionable traceability: The requirements and process hierarchies provide structured, auditable paths from foundational human-centered principles to operational system behavior.

Notable implementation challenges include interdisciplinary fragmentation, lack of design rigor, organizational inertia, regulatory gaps, and scalability limitations; corresponding recommendations focus on integrated processes, standardized metrics, cross-training, and hybrid human-machine deployment models (Xu et al., 2023).

7. Future Directions

Priority areas for advancing hHCAI and its host iSTS framework include:

• Empirical, multi-level deployments: Domain trials in smart cities, aviation, and healthcare embedding co-learning cycles and refined governance.
• Theoretical deepening: Investigating co-evolution dynamics, sociotechnical drift, and resilient system architectures.
• Hybrid methodologies: Integrating cognitive work analysis, joint cognitive systems, resilience engineering, and macroergonomics into unified toolkits.
• International standards: Pursuing new ISO/IEEE standards for human-centered AI, and expanding interdisciplinary HCAI research and education (Xu et al., 2024, Xu et al., 2023).

Through systematically linking design philosophy, stakeholder needs, technical process, and governance, hierarchical HCAI extends HCAI practice to meet the challenges of scalable, transparent, and sustainable AI in complex sociotechnical environments.