Smart sustainability through system satisfaction: Tailored preference elicitation for energy-saving recommenders

People can adopt many different energy-saving measures, but how can they be encouraged to take action? Recommender systems could offer a solution, but how recommender systems are used and perceived will depend on the level of knowledge people have regarding energy-saving measures. We test an energysaving recommender system that uses Multi-Attribute Utility Theory (MAUT) to recommend energysaving measures to its users. Across four experiments we test nine different preference elicitation methods for this system, and demonstrate that users' system satisfaction with each of these interfaces depends on whether they are an expert on energy-saving or a novice. Moreover, we show that system satisfaction is a driver of behavioral outcomes. In effect, a suitable preference elicitation method not only makes users more satisfied with the system; it also entices them to choose more measures with higher average savings, and makes them more satisfied with their choices as well.