The Art of Gift-Giving with Limited Preference Data: How Fashion Recommender Systems Can Help

Sharareh Alipour, Sina Elahimanesh, Aliakbar Ghayouri Sales, Iman Mohammadi, Parimehr Morassafar, Seyed Parsa Neshaei

Abstract: Gift shopping can be challenging due to the limited prior knowledge of the recipient’s preferences, leading to after-purchase regret. The effectiveness of Fashion Recommender Systems (FRS) in the context of gift purchases with limited preference data remains under-explored. We considered a gift-buying scenario and conducted an experiment with 192 pairs of participants to compare FRS versus humans in recommending fashion gifts to buyers. We find both FRS and humans score >50% correctness in recommending the right gift, even without direct interaction with gift-givers or recipients. Although the buyers know the gift receivers directly, they lead to less accuracy. Additionally, we identify gender-based differences in the recommendations. We also embed our scenario into a smartphone application. Our findings investigate the potential of FRS in cold-start scenarios with limited data and unavailable human assistance while highlighting the risks of using FRS for in-store gift purchases.