A Cascaded Human-in-the-Loop Governance Architecture With Feedback-Driven Escalation Optimization for Ethical Artificial Intelligence-Augmented Workforce Systems

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).

The increasing use of intelligent automation in human-focused workplaces has intensified the need for governance mechanisms that allow the preservation of human power without compromising operational effectiveness.Traditional automation systems are often designed using static oversight models that are not responsive enough to contextual risk, changing system behavior, and dynamic human-AI interaction patterns.A cascaded human-in-the-loop governance architecture is one of the proposed architectures in this paper that incorporates adaptive oversight into intelligent automation processes.The proposed framework formalizes governance intensity as a dynamical system property, which is organized around five coordinated layers, namely: automation execution monitoring, probabilistic risk stratification with uncertainty-sensitive anomaly detection, ethical compliance validation, adaptive human escalation orchestration with configurable risk thresholds, and a feedback-based learning module, which augments the escalation policies over time through human correction as a feedback mechanism.As opposed to understanding governance as an ex post compliance role, the architecture realizes structured human-AI cooperation to maintain accountability, transparency, and labor agency in high-risk decision-making situations.Improvements in decision accuracy, compliance adherence, and risk detection sensitivity over a base of 15,000 simulated workforce workflow decisions are statistically significant and do not incur suboptimal productivity trade-offs.Longitudinal analysis also shows that unnecessary escalations are gradually reduced through feedback-induced optimization of the system, which proves that the system has the ability to achieve adaptive control without reducing human control.The results contribute to the design of intelligent systems that are people-centered, which can be achieved through the establishment of adaptive governance as a structural principle of intelligent system design in order to guarantee ethical, trustworthy, and sustainable intelligent automation across various fields of operation.