Quantifying and encoding occupants' preferences as an objective function for\nthe tactical decision making of autonomous vehicles is a challenging task. This\npaper presents a low-complexity approach for lane-change initiation and\nplanning to facilitate highly automated driving on freeways. Conditions under\nwhich human drivers find different manoeuvres desirable are learned from\nnaturalistic driving data, eliminating the need for an engineered objective\nfunction and incorporation of expert knowledge in form of rules. Motion\nplanning is formulated as a finite-horizon optimisation problem with safety\nconstraints. It is shown that the decision model can replicate human drivers'\ndiscretionary lane-change decisions with up to 92% accuracy. Further proof of\nconcept simulation of an overtaking manoeuvre is shown, whereby the actions of\nthe simulated vehicle are logged while the dynamic environment evolves as per\nground truth data recordings.\n