Consider the following situation. You are asked to build a recommendation algorithm for a social media company. The algorithm will use AI to review user interactions with the system (e.g., text, images, GIFs, videos, likes, reposts). These interactions will be used to develop a user profile. The algorithm will recommend products to the user based on user profile characteristics. These characteristics include (but are not limited to) gender, geographic location, age, interests (e.g. sports, movies, books, travel, concerts, etc.), socio-economic class, job profile, relationships, health, and more. 

What harm might this recommendation system cause?