ESTRO 2024 - Abstract Book

S2858

Interdisciplinary - Health economics & health services research

ESTRO 2024

Results:

In total 211 healthcare providers completed the survey. Respondents noted high trust in AI ( X̄ = 3.59, s = 0.41, on a 5-point Likert scale). Furthermore, no distinct difference was found between the professions with regard to associated risks, perceived AI performance, perceived control and desire of explainability. When the software was externally developed, respondents had a higher desire for explainable AI than when the AI was internally developed ( X̄ external= 4.13, X̄ internal = 3.28, p = 0.0384). The follow up analysis aligned with the findings of our survey. Participants of the workshop rated average trust in Dutch RT centers 6.3 on a scale from 1 to 10. Furthermore, most important characteristics of trustworthy AI were validation, explainability, and reproducibility (Figure 1). Most mentioned bottlenecks to the implementation of AI were trust, the black box characteristics of AI, and education and knowledge (Figure 2).