Generative AI as a recursive observation system: a second-order cybernetic perspective on cognitive autonomy and human-centered governance

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Purpose This paper aims to analyze generative artificial intelligence ecosystems from the perspective of second-order cybernetics. It elucidates how AI-mediated environments transform human observation, cognition, and the construction of reality. By revisiting Heinz von Foerster's concepts of observer-dependency, reflexivity, non-trivial systems, and the ethical imperative, the study formulates a human-centered governance framework to address emerging epistemic and socio-technical complexities in AI-mediated societies. Design/methodology/approach The study employs a systems-oriented conceptual methodology grounded in second-order cybernetics, socio-technical systems theory, and contemporary AI governance literature. It integrates concepts from cybernetics, cognition, algorithmic mediation, reflexive governance, and generative AI into a unified framework. This framework conceptualizes generative AI ecosystems as recursive socio-technical systems shaped by ongoing interactions among human observers, intelligent systems, institutions, and digitally mediated realities. Findings The study demonstrates that generative AI systems serve as epistemic and cognitive mediators rather than solely as computational tools. Through algorithmic filtering, generative content production, adaptive personalization, and recursive informational feedback, these systems actively influence interpretation, meaning construction, and human cognition. Existing first-order governance approaches are limited, as they primarily focus on technical regulation and accountability while neglecting reflexivity, observer-dependency, and cognitive autonomy. The proposed framework underscores the significance of reflexive governance, interpretive plurality, cognitive resilience, and governance-by-design in preserving meaningful human agency within AI-mediated environments. Originality/value This study advances second-order cybernetics in contemporary generative AI governance discourse by conceptualizing AI ecosystems as recursive socio-technical observation systems that integrate cognition, reflexivity, and human-centered governance.