Orionis GroupTechnology in Society · 2026 · Digital innovation special issue
Misalignment without misbehavior
Cross-level institutional alignment in AI-enabled sociotechnical systems
How can technically adequate or locally compliant operation contribute to divergence from a broader institutional objective? This paper examines the cross-level condition and the mechanisms that can connect workflow success with institutional change.
Analysis requires a specified objective reference, evidence of local adequacy, institutional divergence and a credible causal linkage between them.
Research manuscript
Control Erosion in AI-Enabled Institutions
How Successful Automation Can Weaken Human Oversight
This work examines declining practical capacity to understand, redirect and recover from or exit an AI-mediated process. It represents control through observability, steerability and reversibility and develops candidate indicators and governance questions.
The analysis uses structured scenarios and bounded evidence from existing studies. Empirical validation of the proposed diagnostic architecture remains a research task.
Research manuscript
Quiet Failure in AI Governance
A Trajectory-Level Blind Spot in Alignment Monitoring
Quiet Failure examines materially concerning evidence of departure from an institutional objective while behavior stays within expected bounds and violation channels remain quiet.
The framework asks what institutions can observe responsibly, how measurements should be calibrated and when evidence can support classification. Synthetic results establish feasibility within the specified benchmark. External institutional validation is a further research stage.
Earlier publication
Jean Y Djamen (2018). Exploring a new model of public administration in a digital age: automation, digitization and transparency. AAPA 2018, Atlantis Press.