Automatically optimizing a predictive dynamic bending light function of a vehicle lighting system comprises configuring a bending light control unit for controlling the lighting system with initial bending light control parameter values as control parameter values to be used, configuring a classification unit for automatically classifying a performance of the bending light control unit into a desired and at least one further performance class, depending on output values of the control unit, configuring a control parameter optimization unit for ascertaining updated control parameter values depending on input values of the control unit and classifications of the performance assigned to them, as well as acquiring driving trajectory parameter values as the input values during a journey, ascertaining output values of the control unit, automatically classifying the performance depending on the ascertained output values, ascertaining updated control parameter values and adapting the control parameter values to be used to the updated control parameter values, wherein the ascertainment of updated control parameter values comprises the application of a genetic algorithm with which a frequency of a classification of the performance into the desired performance class is increased.