Vehicle-to-everything (V2X) technologies enable automated vehicles (AVs) to detect hazards beyond the driver's field of view, thereby enhancing automated decision-making and mitigating potential accidents. However, drivers monitoring partial AVs may misinterpret the system's actions, resulting in discomfort, mistrust, and unsafe driver-initiated interventions. While prior research has primarily examined the performance of such interventions, limited attention has been given to their underlying causes. To address this gap, this study develops a model to quantify factors influencing drivers’ intervention tendencies in V2X-supported AVs. An online survey was conducted using video simulations of automated driving scenarios that varied in automated driving style, hazard type, and driving environment. Responses from 131 participants were analyzed using partial least squares structural equation modeling. The results identified three key intrinsic traits influencing intervention tendency: Driving Confidence (DC), Driving Style (DS), and Initial Trust in Automation (ITA). DC and aggressive DS were positively associated with intervention tendency, whereas ITA has a negative effect. Two psychological factors – Real-time Trust in Automation (RTA) and Risk Perception (RP) – were found to mediate these relationships. Additionally, the effects of RP and RTA were moderated by automated driving style and driving environment. These findings highlight the critical role of driver traits, risk perception, and trust dynamics in driver-initiated interventions, offering both theoretical and practical insights for AV design and driver training programs to reduce inappropriate interventions and enhance overall safety.