Optimizing Interactive Mobility Interfaces with Human Feedback: Multi Objective Bayesian Optimization in Practice

Mark Colley

Abstract: This course introduces Human in the Loop Multi Objective Bayesian Optimization as a method for improving interactive systems when outcomes are subjective, noisy, and competing. The focus is on real HCI settings where designers must balance measures such as trust, perceived safety, comfort, cognitive load, usability, and efficiency. Using case studies from automated vehicles, external human machine interfaces, in vehicle assistants, and accessibility focused VR systems, the course shows how to formalize design parameters, collect human feedback during iterative studies, build surrogate models, and select new candidates through acquisition strategies under limited evaluation budgets. Participants will inspect the full optimization pipeline rather than using it as a black box, including search space design, objective scaling, Pareto fronts, uncertainty handling, and stopping decisions. The course also covers practical failure cases in human centered optimization, including inconsistent raters, study fatigue, context effects, and non transfer across scenarios, and discusses when Bayesian optimization is useful and when simpler methods are the better choice.

 All speakers  Program