Human-in-the-Loop-for-Artificial-Intelligence -Avo
The chapter examines the declining role of human oversight in increasingly autonomous AI systems and argues that the traditional notion of Human‑in‑the‑Loop (HITL) is being co‑opted as a form of AI ethics‑washing. It traces HITL’s origins in military human‑computer interaction, outlines theoretical foundations of human agency (Bandura’s four‑component model) and contrasts them with emerging notions of AI agency and multi‑agent systems. A comprehensive review of global policy documents shows divergent terminology (HITL, HOTL, HOOTL, HIC) and inconsistent accountability provisions. The authors introduce a conceptual framework that evaluates HITL credibility along three diagnostic pillars – power, trust, and complexity – and demonstrate how shifts in these dimensions enable ethics‑washing. Real‑world incidents such as Boeing’s MCAS failure, Tesla autopilot disengagement, predictive policing bias, and AI‑driven manipulation illustrate the consequences of inadequate human agency. The chapter culminates in a governance model that mandates transparent mapping of human authority, trust‑building mechanisms, and complexity management to prevent ethics‑washing, and calls for globally harmonized norms for human agency across the AI lifecycle.
Topics
From Military Roots to Modern Ambiguity: The Evolving Definition of HITL
Chronicles the historical emergence of Human‑in‑the‑Loop from 1940s error‑detector roles in defense systems to its present‑day, loosely‑defined usage across AI governance.