This viewpoint article makes a compelling case that human-in-the-loop AI 'often collapses into superficial reviews and cursory approvals due to time pressures, automation bias, and institutional constraints, making the oversight symbolic rather than substantive'.
CITATION: Viewpoint. Rawan Abulibdeh et al.
The Lancet Volume 407, Issue 10545 p2340-2344 June 06, 2026
Who's really in the loop? Rethinking oversight in AI-assisted health care
SUMMARY
Human-in-the-loop oversight is widely invoked as a safeguard against potential harm from artificial intelligence (AI) used in health care, yet it functions more as symbolic reassurance than substantive protection. We argue that human-in-the-loop fails for three interconnected reasons: AI used in health care can amplify existing structural inequities at unprecedented scale, intersectional harms elude detection by oversight models premised on neutral singular reviewers, and clinicians operate under constraints that preclude meaningful interrogation of algorithmic outputs. Drawing on actor–network theory, feminist epistemology, and political philosopher Iris Marion Young's social connection model of justice, we show that current governance individualises responsibility while obscuring institutional complicity. We propose three pathways towards more substantive accountability: co-reasoning frameworks that position AI as one voice in clinical deliberation, community-owned governance with authority to suspend harmful systems, and institutional liability structures that redistribute responsibility from clinicians to the organisations that design and deploy these tools.
SELECTED EXTRACTS
'Studies show that clinicians often defer to algorithmic outputs even when those outputs conflict with other available information — a pattern of automation bias that reflects cognitive as much as structural pressures.'
'In many low-income and middle-income countries, clinicians care for extraordinarily large patient loads with constrained time and diagnostic resources. In those settings, oversight models premised on continuous human availability are structurally unrealistic and possibly risk-enhancing.'
HIFA profile: Neil Pakenham-Walsh is coordinator of HIFA (Healthcare Information For All), a global health community that brings all stakeholders together around the shared goal of universal access to reliable healthcare information. HIFA has 20,000 members in 180 countries, interacting in four languages and representing all parts of the global evidence ecosystem. HIFA is administered by Global Healthcare Information Network, a UK-based nonprofit in official relations with the World Health Organization. Email: neil@hifa.org