Examining the user evaluation of multi-list recommender interfaces in the context of healthy recipe choices

Multi-list recommender systems have become widespread in entertainment and e-commerce applications. Yet, extensive user evaluation research is missing. Since most content is optimized toward a user’s current preferences, this may be problematic in recommender domains that involve behavioral change, such as food recommender systems for healthier food intake. We investigate the merits of multi-list recommendation in the context of internet-sourced recipes. We compile lists that adhere to varying food goals in a multi-list interface, examining whether multi-list interfaces and personalized explanations support healthier food choices. We examine the user evaluation (i.e., diversity, understandability, choice difficulty and satisfaction) of a multi-list recommender interface, linking choice behavior to evaluation aspects through the user experience framework.

We present two studies, based on (1) similar-item retrieval and (2) knowledge-based recommendation. Study 1 (N = 366) compared single-list (5 recipes) and multi-list recommenders (25 recipes; presented with or without explanations). Study 2 (N = 164) compared single-list and multi-list food recommenders with similar set sizes and varied whether presented explanations were personalized. Multi-list interfaces were perceived as more diverse and understandable than single-list interfaces, while results for choice difficulty and satisfaction were mixed. Moreover, multi-list interfaces triggered changes in food choices, which tended to be unhealthier, but also more goal based.