As vehicles become more automated in the pursuit of comfort and safety, the human-machine interface must adapt to accommodate the intent of both driver and automation. The concept of haptic shared control steering allows lateral control of the vehicle to be continuously shared between driver and automation via the steering wheel. The automation imparts a torque on the steering wheel which serves to both guide the vehicle and inform the driver of the automation's intent without impairing the driver's ability to steer the vehicle. The combination of haptic feedback and seamless transition of vehicle control makes the use of automation more intuitive. This paper describes a Nonlinear Model Predictive Control (NMPC) approach for controlling the strength of the automation torque with the aim of optimizing vehicle motion. NMPC allows for intuitive tuning and customization of the vehicle steering performance through the NMPC parameterization. Results on a vehicle in real driving scenarios show that the proposed control is able to decrease the lateral jerk, a common measure of passenger comfort, compared to the current state of the art by up to 50% while maintaining the same or similar levels of path tracking performance.