A haptic shared control implementation for car steering is developed using the four design choices architecture guidelines. Nowadays, haptic shared control systems are often tuned heuristically and there is little knowledge on how to set the design parameters. In this research a novel driver model was used to generate human-compatible references for the haptic controller with a moderate amount of curve-cutting. In a driving simulator study with different levels of haptic support (LoHS) it is found that driver acceptance increases with higher LoHS values up to 1. In this implementation a higher LoHS equates to stronger steering angle feedforward. Objective driver acceptance is measured by analyzing the torque conflict data and the total conflict is found to decrease by 70% for the highest LoHS condition compared to the lowest setting. This result is verified with higher subjective ratings of the haptic support as the LoHS is increased. The required driver steering effort decreases significantly for higher LoHS values with a reduction of 81% for the highest LoHS condition compared to manual driving. Likewise a small decrease in the steering wheel reversal rate is found as the haptic support is increased. Under high LoHS values the remaining torque conflict is found in the curve entry and exit phases. The addition of individualised prepositioning into the reference has the potential to reduce the remaining conflict. The findings suggest that choosing the right LoHS is of great importance in achieving high acceptance and should thus be the first parameter to be tuned for a haptic controller of the four design choices architecture type.