Human-centered design of intelligent DSS in medicine could be achieved through lower automation levels in
the stage of decision-making. However, reduced automation may affect users differently based on expertise.
The objective of our research was to explore how expertise and the use of various automation patterns in XAI
systems for medical diagnosis affect user performance and UX. We conducted a between-subjects experiment
(N = 21) with medical novices and experts. Participants interacted with two DSS differing in automation levels.
Performance (accuracy, confidence) and automation-related UX (Subjective information processing awareness,
perceived trustworthiness, diagnosticity) were assessed. Results suggest that only novices tend to over-rely on
lower automation. In contrast, experts did not exhibit this tendency, but their subjective information processing
awareness increased. Our findings indicate that reducing automation levels to enhance human-AI integration
may not consistently improve performance or UX. Designers should consider user expertise and context when
developing DSS with lower automation levels.