As human-AI collaboration continues in radiology, should trust in AI be assumed, demanded, or earned? So asks Jan Beger, GE HealthCare's global head of AI advocacy, in a LinkedIn post 8 August.
Trust in digital human-AI collaboration depends on three key levels working together in a dynamic relationship, he explains. This dynamic multi-level relationship will be affected by perception of AI trustworthiness on the individual level, teams-level dynamics, and various roles at the organizational level, as illustrated in the figure below.
Essi Janhunen, Tuuli Toivikko, Kirsimarja Blomqvist, Dominik Siemon. Trust in Digital Human-AI Team Collaboration: A Systematic Review. AMCIS 2024 Proceedings. 2024. Via Jan Beger, LinkedIn.
"This figure shows how trust in digital human-AI collaboration works across different levels and focuses on different targets like AI, human teammates, and the organization," Beger explained in his post, adding that trust isn't just an individual feeling.
"It moves across levels, from personal beliefs to team behavior to organizational policies, and it depends on who or what is being trusted," he wrote. "Whether trust holds up in a digital human-AI collaboration depends on all of that working together."
In a 30 July editorial published in the European Journal of Radiology Artificial Intelligence, Beger addressed rethinking trust in the age of medical AI. "AI tools increasingly assist in diagnosis, triage, treatment planning, and risk prediction. And while responsibility formally still rests with the clinician, the reality is often less clear."
The role of the radiologist is also evolving, and fast, he continued.
"The radiologist is no longer just a reader of scans, but a diagnostic strategist, data integrator, and clinical collaborator -- working at the intersection of precision medicine, digital infrastructure, and clinical care," he said, noting that this expanded role demands more than clinical acumen, AI literacy, systems-level thinking, ethical awareness, and leadership across multidisciplinary teams.
Moreover, "the question isn’t whether we should trust AI. It’s whether AI is designed in a way that deserves trust," Beger wrote.
Read the complete journal article here.
