Abstract
Multistakeholder Recommender Systems (MRSs) mediate possibly contrasting goals of the parties involved in the recommendation process. For example, in the tourism domain an MRS could be designed to balance the sustainable goal of a destination to reduce the overcrowding of popular points of interest, while still satisfying the preferences of tourists who would like to visit popular places. The conflicting nature of these goals raises the question of whether, by means of recommendations, it is possible to achieve better outcomes for both stakeholders, i.e., the tourists and the destination. To answer this question, we simulate the impact of an MRS on tourists’ choices. We focus on a class of MRSs that linearly balance tourists and destination objectives. This is the prevalent balancing approach to date. We assume a realistic scenario where tourists have limited knowledge of the points of interest catalogue and the MRS has little information about tourists’ preferences. Simulations of tourist-system interactions in crowded cities show that this MRS can guide tourists to choose more appropriate places to visit, benefiting both them and the destination. We refer to this as a positive-sum impact. However, a proper balance is crucial: putting under- or over-emphasis on sustainable points of interest in the recommendation lists can be detrimental to both destination sustainability and tourists’ satisfaction. Moreover, with simulations, we show that even non-personalised popularity-driven recommendations, when combined with sustainable recommendations, can have a (less pronounced) positive impact. This suggests that even a simple non-personalised MRS may be useful in urban tourism, if properly used.