Recently, many studies have been conducted on increasing the automation level in cars to achieve safer and more efficient transportation. The increased automation level creates room for the drivers to shift their attention to non-driving related activities. However, there are cases that cannot be handled by automation where a driver should take over the control. This pilot study investigates a paradigm for keeping the drivers' situation-awareness active during autonomous driving by utilizing a social robot system, NAMIDA. NAMIDA is an interface consisting of three sociable driving agents that can interact with the driver through eye-gaze behaviors. We analyzed the effectiveness of NAMIDA on maintaining the drivers' attention to the road, by evaluating the response time of the drivers to a critical situation on the road. An experiment consisting of a take over scenario was conducted in a dynamic driving simulator. The results showed that existence of NAMIDA significantly reduced the response time of the drivers. However, surprisingly, NAMIDA without eye-gaze behaviors was more effective in reducing the response time than NAMIDA with eye-gaze behaviors. Additionally, the results revealed better subjective impressions for NAMIDA with eye-gaze behaviors behaviors.