As automated vehicles (AVs) become increasingly integrated into real-world traffic, understanding drivers’ trust dynamics and intervention tendencies in potential hazard scenarios is critical for ensuring safe and effective human-AV collaboration. This study examines how drivers calibrate trust and adjust their intervention behavior when repeatedly exposed to ghost probe events – sudden emergences of vulnerable road users from occluded areas. A driving simulator experiment was conducted to investigate drivers’ responses to AVs exhibiting defensive (decelerating and stopping) and assertive (maintaining speed without collision) driving styles in potential hazard scenarios. Twelve participants experienced five consecutive ghost probe events, with only the fourth involving an actual pedestrian running into the road. Trust levels and intervention tendencies were assessed after each event using self-reported questionnaires, followed by a short interview. Results indicate that in the first three events, the defensive AV style facilitated progressive trust stabilization and a gradual decline in intervention tendency, whereas the assertive AV style initially led to trust erosion, followed by adaptive recalibration and a subsequent decrease in intervention tendency. The fourth event induced increased trust in defensive AVs and further reduced intervention tendency, while it diminished trust in assertive AVs yet also lowered intervention tendency. By the fifth event, both trust levels and intervention tendencies returned to levels observed in the third event. These findings offer insights into trust adaptation mechanisms and driver intervention behaviors in AV interactions, informing the design of AV systems to optimize trust calibration, reduce unnecessary interventions, and enhance overall safety in mixed-traffic environments.