Artificial intelligence is becoming progressively embedded within institutional processes that support perception, evidence retrieval, classification, modelling, interpretation, scenario generation and decision support. This development extends a much longer history in which institutions have relied upon external cognitive technologies to preserve memory, organise information and expand analytical capacity. Contemporary AI systems, however, increasingly participate in intermediate cognitive transformations that shape how institutional problems are represented before formal decisions occur. Their significance therefore cannot be understood adequately through model performance, automation rates or the continued formal presence of human decision-makers alone.
This Working Paper develops the Human–AI Institutional Collaboration Architecture (HAICA) as a middle-range framework for analysing how human and artificial cognitive capabilities become distributed, connected and coordinated within institutions. Building upon the Institutional Cognition Conceptual Architecture and the Institutional Cognitive Architecture Model, HAICA introduces Hybrid Institutional Cognition as the distributed performance of institutional cognitive functions through configurations combining human actors, artificial cognitive systems, organisational procedures and information infrastructures. The framework uses the Human–AI Cognitive Configuration (HACC) as its principal operational unit, allowing specific arrangements of human and artificial participation to be reconstructed and compared without assuming a permanent division of cognitive labour between people and machines.
HAICA develops Cognitive Function Decomposition and Cognitive Function Allocation as mechanisms for determining where artificial participation becomes institutionally relevant, while Human–AI Cognitive Complementarity examines the conditions under which differentiated capabilities can become mutually corrective rather than merely additive. Hybrid Cognitive Interfaces and Human–AI Cognitive Translation describe the processes through which differently structured representations become mutually usable, and Hybrid Epistemic Traceability preserves the cognitive lineage through which evidence, computational transformations and human interpretation influence institutional judgement. Particular attention is given to the distinction between Cognitive Authority and Decision Authority, demonstrating that formal human control can remain intact while effective epistemic influence shifts substantially towards artificial systems.
The framework therefore replaces procedural notions of human presence with Meaningful Human Judgement, understood as the condition in which authorised human actors retain sufficient cognitive access, contextual understanding, time, alternative information and institutional authority to understand, challenge and alter artificial contributions when doing so is consequential. HAICA also examines pathways through which initially productive augmentation may deteriorate, including Artificial Cognitive Deference, Augmentation Overload, Institutional Cognitive Deskilling, Cognitive Dependency and Hybrid Institutional Fragility. These vulnerabilities are balanced by Human–AI Cognitive Resilience, Institutional Cognitive Autonomy and Hybrid Cognitive Security, which describe the capacities required to preserve independent judgement, alternative cognitive pathways and the integrity of hybrid institutional reasoning.
The architecture is explicitly adaptive. Human–AI Feedback Architecture provides evidence concerning the behaviour of hybrid configurations; Hybrid Institutional Learning incorporates that evidence into organisational knowledge and practice; Hybrid Institutional Metacognition makes the distribution of human and artificial cognition itself an object of institutional understanding; and Adaptive Cognitive Allocation allows functions to be reassigned as technological capabilities, institutional expertise and risk conditions change. HAICA therefore does not prescribe a fixed boundary between human and artificial cognition. Its central scientific contribution is to establish hybrid cognitive architecture as the relevant object through which institutional AI should be evaluated, shifting research from the question of what artificial intelligence can automate towards the question of how institutions can deliberately organise, observe and govern evolving distributions of human and artificial cognition while preserving epistemic accountability, resilience and legitimate authority.
