Physicians Assign Themselves More Responsibility Than the General Population when AI Fails in Healthcare: A Swedish Cross-Sectional Survey

Abstract Background: The integration of Artificial Intelligence (AI) into clinical decision-making creates uncertainty regarding responsibility and accountability when adverse outcomes occur. Despite this, data regarding stakeholder perspectives remain scarce, especially quantitative data comparing physician and public views on responsibility allocation and transparency. This study aims to evaluate and compare how physicians and the general population in Sweden attribute responsibility for negative medical outcomes when AI is used to support healthcare decisions in primary care, as well as their views on the necessity of transparently disclosing AI use to patients.Methods: A nationwide, cross-sectional, mixed-methods survey was conducted in Sweden between February and May 2025. The survey targeted two distinct cohorts (N = 500 per group): licensed physicians under 70 years, sampled from the National Board of Health and Welfare register, and adults aged 18 or older sampled from the Swedish population registry. The survey used case vignettes as background for multiple-choice questions, with complementary open-ended free-text responses. The data was evaluated using comparative statistical analysis and reflexive thematic analysis.Results: Physicians were significantly more likely than the general population to assign responsibility to the physician (92\% vs. 74\%) and to favor a shared responsibility model involving the physician (48\% vs. 35\%). In contrast, the general population more frequently selected non-physician actor combinations (26\% vs. 8\%) and assigned greater responsibility to the patient in a scenario involving direct patient-AI interaction without a physician involved. Neither group broadly assigned responsibility to the AI itself (1.3\% physicians vs. 10\% general population). The reflexive thematic analysis reinforced these results, identifying a shared accountability framework where all actors share responsibility, but where the physician serves as the ultimate safeguard. A majority of the general population (68\%) affirmed the patient's right to always know if AI was used, in contrast to only 23\% of physicians who exhibited a polarised distribution. Conclusion: Respondents endorse a shared responsibility model among actors involved in implementing and using AI for medical decisions. The physician is viewed as the essential human-in-the-loop, bearing ultimate accountability for adverse outcomes of AI-assisted medical decisions according to a vast majority of respondents. Divergent perspectives regarding transparency emphasise the need to clarify informed consent standards for AI deployment in healthcare.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11280050/v1
Primary Topic
Artificial Intelligence in Healthcare and Education
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Physicians Assign Themselves More Responsibility Than the General Population when AI Fails in Healthcare: A Swedish Cross-Sectional Survey

Ronny Gunnarsson, Anna Moberg, Carl Wikberg, Peter Nymberg et al.
2 links
Research Square
Artificial Intelligence in Healthcare and Education
preprint

Physicians Assign Themselves More Responsibility Than the General Population when AI Fails in Healthcare: A Swedish Cross-Sectional Survey

Ronny Gunnarsson, Anna Moberg, Carl Wikberg, Peter Nymberg, Sofia Thunberg, Rasmus Arvidsson, Jonathan Widén, Charlotte Blease, Pär-Daniel Sundvall, David Sundemo, Daniel Andersson
preprint en

Abstract

Abstract Background: The integration of Artificial Intelligence (AI) into clinical decision-making creates uncertainty regarding responsibility and accountability when adverse outcomes occur. Despite this, data regarding stakeholder perspectives remain scarce, especially quantitative data comparing physician and public views on responsibility allocation and transparency. This study aims to evaluate and compare how physicians and the general population in Sweden attribute responsibility for negative medical outcomes when AI is used to support healthcare decisions in primary care, as well as their views on the necessity of transparently disclosing AI use to patients.Methods: A nationwide, cross-sectional, mixed-methods survey was conducted in Sweden between February and May 2025. The survey targeted two distinct cohorts (N = 500 per group): licensed physicians under 70 years, sampled from the National Board of Health and Welfare register, and adults aged 18 or older sampled from the Swedish population registry. The survey used case vignettes as background for multiple-choice questions, with complementary open-ended free-text responses. The data was evaluated using comparative statistical analysis and reflexive thematic analysis.Results: Physicians were significantly more likely than the general population to assign responsibility to the physician (92\% vs. 74\%) and to favor a shared responsibility model involving the physician (48\% vs. 35\%). In contrast, the general population more frequently selected non-physician actor combinations (26\% vs. 8\%) and assigned greater responsibility to the patient in a scenario involving direct patient-AI interaction without a physician involved. Neither group broadly assigned responsibility to the AI itself (1.3\% physicians vs. 10\% general population). The reflexive thematic analysis reinforced these results, identifying a shared accountability framework where all actors share responsibility, but where the physician serves as the ultimate safeguard. A majority of the general population (68\%) affirmed the patient's right to always know if AI was used, in contrast to only 23\% of physicians who exhibited a polarised distribution. Conclusion: Respondents endorse a shared responsibility model among actors involved in implementing and using AI for medical decisions. The physician is viewed as the essential human-in-the-loop, bearing ultimate accountability for adverse outcomes of AI-assisted medical decisions according to a vast majority of respondents. Divergent perspectives regarding transparency emphasise the need to clarify informed consent standards for AI deployment in healthcare.

Research Square
Linköping University (SE), Uppsala University (SE), Lund University (SE), Region Västra Götaland (SE), University of Gothenburg (SE)
Good health and well-being
Artificial Intelligence in Healthcare and Education
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon